Johnson Shane M 201008 PHD

COMPUTATIONAL MODELING, STOCHASTIC AND
EXPERIMENTAL ANALYSIS WITH THERMOELASTIC

STRESS ANALYSIS FOR FIBER REINFORCED
POLYMERIC COMPOSITE MATERIAL SYSTEMS
A Thesis
Presented to
The Academic Faculty
by
Shane M. Johnson
In Partial Fulfillment
of the Requirements for the Degree
Doctor of Philosophy in the
School of Civil and Environmental Engineering
Georgia Institute of Technology

August 2010
COMPUTATIONAL MODELING, STOCHASTIC AND

EXPERIMENTAL ANALYSIS WITH THERMOELASTIC
STRESS ANALYSIS FOR FIBER REINFORCED
POLYMERIC COMPOSITE MATERIAL SYSTEMS
Approved by:
Professor Rami Haj-Ali, Advisor
Department of Civil Engineering
Professor Abdul-Hamid Zureick

Professor Kenneth Will

Professor Andrew Makeev

Department of Aerospace Engineering
Professor Donald White

Date Approved: April 2010
ACKNOWLEDGEMENTS
My thanks and appreciation go to Dr. Rami Haj-Ali, my advisor, for his guidance,
encouragement, persistence, patience and expert advice. Interaction with him has
inspired my love of research and has encouraged me to seek new and exciting challenges. My graditude also goes to my colleagues Rani El-Hajjar, Bo-Siou Wei, and
Hee-Sun Kim. Thanks are also in order to Bradley Boyce of Stress Photonics Inc.
for his expert advice and valuable technical assistance in the field of Infrared Thermography. Thanks to Erian Armanios for opening his composites manufacturing lab
to my research group, and thanks to Xinyuan Tan for guiding me to manufacturing
quality composites.
iii
TABLE OF CONTENTS
ACKNOWLEDGEMENTS . . . . . . . . . . . . . . . . . . . . . . . . . . . .
iii
LIST OF TABLES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
vii
LIST OF FIGURES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
viii
LIST OF SYMBOLS OR ABBREVIATIONS . . . . . . . . . . . . . . . . . .
xiv
SUMMARY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
xiv
INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.1 General . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.1.1
Fiber Reinforced Polymeric Materials . . . . . . . . . . . . .
1.1.2
Thermography . . . . . . . . . . . . . . . . . . . . . . . . .
1.2 Experimental Studies . . . . . . . . . . . . . . . . . . . . . . . . . .
1.2.1
II
Quantitative Thermography and Thermoelastic Stress Analysis in Composites . . . . . . . . . . . . . . . . . . . . . . .
1.2.2
TSA for Damage in FRP Composites . . . . . . . . . . . . .
1.2.3
Experimental Studies on Jute Hybrid Bio-Composites . . .
1.2.4
Damage in FRP Lap Joints . . . . . . . . . . . . . . . . . .
10
1.3 Analytical and Numerical Studies . . . . . . . . . . . . . . . . . . .
12
1.3.1
Nonlinear Anisotropic Micro-macromechanical Modeling . .
12
1.3.2
Analytical and Computational Modeling of FRP Lap Joints
16
1.3.3
Fatigue Damage Models for FRP Composites . . . . . . . .
19
1.4 Objective . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
22
1.5 Present Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
23
THERMOELASTIC STRESS ANALYSIS FOR QUANTITATIVE ANALYSIS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

2.1 Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
26
2.2 Derivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
27
2.3 IR Test Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
33
iv
III
2.4 Validation of TSA Technique on Prepreg and Pultruded Composites
36
2.5 Thermomechanical Calibrations for Various FRP Composites . . .
37
2.6 Validation under a Multi-axial state of stress . . . . . . . . . . . .
42
MECHANICAL BEHAVIOR OF JUTE HYBRID BIO-COMPOSITES
54
3.1 Introduction
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
54
3.2 Jute Hybrid Bio-composite Systems . . . . . . . . . . . . . . . . . .
55
3.3 Nonlinear Constitutive Models for Bio-Composites . . . . . . . . .
57
3.3.1
Anisotropic Deformation Theory (ADT) Based Model . . .
59
3.3.2
Anisotropic Potential Theory Model (APT) . . . . . . . . .
61
3.3.3
New Formulation for the Nine-Cell Micromechanical Model

(Nine-Cell) . . . . . . . . . . . . . . . . . . . . . . . . . . .
64
The HFGMC micromechanical modeling framework . . . . .
72
3.4 Calibration and Verification . . . . . . . . . . . . . . . . . . . . . .
88
3.3.4
IV
THERMOELASTIC STRESS ANALYSIS FOR FAILURE INITIATION

AND PROGRESSION IN COMPOSITE LAP SHEAR JOINTS . . . . . 110
4.1 Introduction
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
4.2 Proposed IR-TSA Testing Set-up . . . . . . . . . . . . . . . . . . . 111

4.3 Proposed IR Testing Approach . . . . . . . . . . . . . . . . . . . . 114
4.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
V
A STOCHASTIC FATIGUE DAMAGE MODEL FOR COMPOSITE SINGLE LAP SHEAR JOINTS BASED ON MARKOV CHAINS AND THERMOELASTIC STRESS ANALYSIS . . . . . . . . . . . . . . . . . . . . . 126
5.1 Introduction
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
5.2 Testing Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129

5.3 Fatigue damage metrics using IR-TSA . . . . . . . . . . . . . . . . 133
5.4 A cumulative damage model using Markov chain theory . . . . . . 138
5.5 TSA results and Stochastic S-N curves . . . . . . . . . . . . . . . . 140
5.6 TSA results and Stochastic S-N curves . . . . . . . . . . . . . . . . 149
VI
CONCLUSION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
6.1 Quantitative Thermoelastic Stress Analysis . . . . . . . . . . . . . 152

6.2 Analytical and Experimental Analyses for Jute Hybrid Bio-Composites152
6.3 Damage Initiation of FRP Lap Joints . . . . . . . . . . . . . . . . . 153
6.4 Stochastic Fatigue of FRP Lap Joints . . . . . . . . . . . . . . . . 155
6.5 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
6.5.1
Fatigue of Lap Joints . . . . . . . . . . . . . . . . . . . . . . 156
6.5.2
Fatigue of 3D Woven Fiber Reinforced Polymeric Material

Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
REFERENCES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
vi
LIST OF TABLES
1
Effective Elastic Properties . . . . . . . . . . . . . . . . . . . . . . . .
36
Description of Jute hybrid bio-composite material system . . . . . . .
55
Nominal V-notch Specimen Dimensions . . . . . . . . . . . . . . . . .
91
Unidirectional elastic material properties of Jute with chopstrand mat
92
Calibrated Ramberg-Osgood Relations and Parameters for Jute / Glass

Chopstrand Mat . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
93
Ultimate material properties of Jute with chopstrand mat . . . . . . .
93
Fiber linear elasetic properties used in the Nine-Cell RUC . . . . . .
94
Nine-Cell RUC Isotropic Nonlinear Matrix Material Properties (RambergOsgood Relations) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
Nine-Cell RUC geometry for equivalent constituent volume fractions .
94
10
Nine-Cell RUC in-plane cell properties for equivalent volume fractions
96
11
Unidirectional material properties of IM7/epoxy composite laminate . 115
12
Unidirectional material properties of S2-glass/epoxy composite laminate132
13
Number of test specimens and applied stresses . . . . . . . . . . . . . 133
14
Number of test specimens and applied stresses . . . . . . . . . . . . . 140
vii
LIST OF FIGURES
1
Schematic cross-sectional view of a layered orthotropic composite with

surface coatings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
28
Schematic of TSA Setup for thermomechanical calibration . . . . . .
34
Setup for Thermo-mechanical calibration . . . . . . . . . . . . . . . .
35
Thermo-mechanical calibration for E-glass/epoxy . . . . . . . . . . .
37
Thermo-mechanical calibration for Uncoated Carbon/Epoxy . . . . .
38
Schematic showing the lay-up for [05 /90/05 ] . . . . . . . . . . . . . .
38
Thermo-mechanical calibration for uncoated S2glass/Epoxy for the

[05 /90/05 ] Lay-up . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
39
Thermo-mechanical calibration for coated S2-glass/epoxy . . . . . . .
40
Thermo-mechanical calibration for quasi isotropic S2-glass/Epoxy . .
41
10
Geometry of notched specimen used to verify the TSA technique and a

very refined quarter-model FE mesh used to model the notched specimen 43
11
TSA images of open hole specimens of Carbon/epoxy . . . . . . . . .
12
Horizontal line interrogation of open-hole geometry for verification of

thermomechanical calibration of un-coated unidirectional carbon/epoxy 45
13

thermomechanical calibration of coated unidirectional carbon/epoxy .
46
Vertical line interrogation of open-hole geometry for verification of

thermomechanical calibration of coated unidirectional carbon/epoxy .
46
14
44
15

thermomechanical calibration of un-coated unidirectional carbon/epoxy 47
16
TSA images of open hole specimens of S2-glass/epoxy . . . . . . . . .
48
17
Horizontal line interrogation of open-hole geometry for verification

of thermomechanical calibration of un-coated S2-glass/epoxy for the
[05 /90/05 ] lay-up . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
49

thermomechanical calibration with epoxy coating of S2-glass/epoxy for
the [05 /90/05 ] lay-up . . . . . . . . . . . . . . . . . . . . . . . . . . .
49
18
viii
19

thermomechanical calibration of un-coated quasi-isotropic S2-glass/epoxy
lay-up . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
20

thermomechanical calibration with epoxy coating of quasi-isotropic S2glass/epoxy lay-up . . . . . . . . . . . . . . . . . . . . . . . . . . . .
51
21

thermomechanical calibration of un-coated S2-glass/epoxy for the [05 /90/05 ]
lay-up . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
22

thermomechanical calibration with epoxy coating of the S2-glass/epoxy
for the [05 /90/05 ] lay-up . . . . . . . . . . . . . . . . . . . . . . . . .
52
23

thermomechanical calibration for un-coated S2-glass/epoxy with a quasiisotropic lay-up . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
24

thermomechanical calibration with epoxy coating for S2-glass/epoxy
with a quasi-isotropic lay-up . . . . . . . . . . . . . . . . . . . . . . .
53
25
Hybrid Jute/Glass composite with chopstrand mat surface layer . . .
56
26
Color similarity algorithms used to identify and segment Jute yarns

from background . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
56
(a) Idealized medium with periodic arrays of long fibers (rectangular

cross-sections) showing a quarter UC (bold line), (b) Nine-cell micromodel of rectangular UC . . . . . . . . . . . . . . . . . . . . . . . . .
65
27
28
29
30
31
Schematic illustration of a unidirectional periodic array in the global

x2 x3 plane of multiphase composite media with its repeating unit-cell
(RUC), defined with respect to its y2 y3 local coordinate system. .
74
General arrangement of cells and their geometry and coordinate systems for the HFGMC-RUC model. Mirrored cells-interfaces are illustrated to enforce the periodic boundary conditions. . . . . . . . . . .
75
One-cell configuration with its neighboring cells showing its two primary feed-forward interfaces along with the location for the stress
integration points. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
82
Overall tangential system of equations for the HFGMC-RUC model

indicating the contribution of cell () to the system and showing a
row of complete equations for this cell. . . . . . . . . . . . . . . . . .
85
ix
32
Jute/glass chopstrand mat system as tested in tension in the fiber

direction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
89
Jute/glass chopstrand mat system as tested in compression in the

transverse fiber direction . . . . . . . . . . . . . . . . . . . . . . . . .
90
Jute/glass chopstrand mat v-notch specimen as tested with the Iosipescu

fixture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
91
Jute/glass chopstrand mat system as tested in shear with the Iosipescu

fixture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
92
36
Nine-Cell RUC geometry for System 1 . . . . . . . . . . . . . . . . .
95
37
Contour map of TSA image scaled to predicted remote axial stress and
contour map of calculated axial stress from an optical strain analysis
method of an open-hole specimen (Specimen 1) of Jute with chopstrand
mat surface layer . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
97
Contour map of TSA image scaled to predicted remote axial stress

and contour map of calculated axial stress from an Digital Image Correlation method of an open-hole specimen (Specimen 2) of Jute with
chopstrand mat surface layer . . . . . . . . . . . . . . . . . . . . . . .
98
33
34
35
38
39
Contour map of predicted remote axial stress of open-hole specimen

Jute with chopstrand mat surface layer showing ADT and APT Models 99
40
Contour map of predicted remote axial stress of open-hole specimen

Jute with chop strand mat surface layer showing Linear and Nine-Cell
Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
41
Axial Stress line interrogation on open-hole specimen for 400lb loading

on Jute with chop strand mat material system showing the prediction
capabilities of the Nine-Cell Model and Linear orthotropic models . . 101
42
Axial Stress line interrogation on open-hole specimen for 400lb loading

on Jute with chop strand mat material system showing the prediction
capabilities of the ADT, APT and Linear orthotropic models . . . . . 102
43
Translaminar Fracture Toughness Test Specimen Schematic . . . . . . 103
44
Results of the Translaminar Fracture Toughness Test on Jute/Glass

Chopstrand using the Eccentrically Loaded Single Edge Notched Tension Specimen . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
45
Remote Stress vs. Remote Axial Strain, Image Correlation at Location

1 used for validation of the ADT, APT, Nine-Cell models, Linear, and
XFEM orthotropic models . . . . . . . . . . . . . . . . . . . . . . . . 106
46
Remote Stress vs. Axial Strain at Open Hole Edge, Image Correlation
at Location 2 used for validation of the ADT, APT, Nine-Cell models,
Linear, and XFEM orthotropic models . . . . . . . . . . . . . . . . . 107
47
Remote Stress vs. Axial Strain, Digital Image Correlation at Location

3 used for validation of the ADT, APT, Nine-Cell models, Linear, and
XFEM orthotropic models . . . . . . . . . . . . . . . . . . . . . . . . 108
48
Remote Stress vs. Axial Strain, Digital Strain Analysis using 3/4 inch
gauge length extensometer over the open hole used for validation of the
ADT, APT, Nine-Cell models, Linear, and XFEM orthotropic models 109
49
Infrared camera and the IR-TSA experimental setup . . . . . . . . . 112
50
Areas of interest and IR Window . . . . . . . . . . . . . . . . . . . . 113
51
Schematic for Expected IR Front View of Lap Joint . . . . . . . . . . 114
52
Single Lap Joint specimen geometry made from IM7 Quasi-isotropic

laminates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
53
Deformed single lap joint . . . . . . . . . . . . . . . . . . . . . . . . . 116
54
Peel and shear stresses in a typical IM7/Epoxy quasi-isotropic single

lap shear joint . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
55
Localized stress intensity at bond edge, FE results showing peel stresses

at bond edge on a deformed IM7/Epoxy quasi-isotropic single lap joint 117
56
Two proposed IR-TSA Test Methods. Method-A: a static load step is

applied, followed by a cyclic loading about this loading level. MethodB: a static load is applied and unloaded followed by cyclic loading at a
pre-specified load and cyclic load amplitude. The next applied static
load level is increased. . . . . . . . . . . . . . . . . . . . . . . . . . . 118
57
Front and side view of the crack front formation in long-bond singlelap shear joint made with woven plates and tested with Method A,
ultimate failure occurs immediately following the 4,450 lb load level.
Signature of crack tip is visible from the side view at the 4,400 lb load
level. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
58
Front and side view of the crack front formation at the bond edge in
long-bond single-lap shear joint made from uni-tape plates tested with
Method A, ultimate failure occurs immediately following the 3,300 lb
load level. Signature of crack tip is visible from the side view at the
3,200 lb load level. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
xi
59
Early stages of crack formation and damage initiation in short-bond

single-lap shear joint made from woven plates tested with procedure B
preserved after a maximum load level of 1700lb. Noticeable changes in
the bond line IR measurement from (c) to (d). . . . . . . . . . . . . . 121
60
Early stages of crack formation and damage initiation in short-bond

single lap-shear joint made with uni-tape plates tested with Method
B and preserved after a maximum load level of 2000lb. Increasing
changes in bondline IR emission indicating damage progression from
(b) to (c) and (c) to (d). . . . . . . . . . . . . . . . . . . . . . . . . 122
61
Early stages of crack formation and damage initiation in long-bond

single-lap shear joint made from woven plates tested with Method B
and preserved after a maximum load level of 3750lb. Increasing changes
in bondline IR emission indicating damage progression from (b) to (c)
and (c) to (d). Noticeable drop in stress concentration from (c) to (d). 122
62
Early stages of crack formation and damage initiation in long-bond

single-lap shear joint made from uni-tape plates tested with Method
B and preserved after a maximum load level of 1900lb. Increasing
changes in bondline IR emission indicating damage progression from
(c) to (d). Noticeable drop in stress concentration from (c) to (d). . . 123
63
Photograph showing micro-cracking in short bond line fabric specimen

#1 (200X) corresponding to IR-TSA measurement in Figure 59. . . . 124
64
Photograph showing disbonding at the adhesive interface in short bondline fabric specimen #4 (200X) . . . . . . . . . . . . . . . . . . . . . 124
65
Photograph showing microcracking in short bond line uni-tape specimen #5 (50X) corresponding to IR-TSA in Figure 60. . . . . . . . . 125
66
Fatigue Test Setup for S2-glass/E733FR Single Lap Joints . . . . . . 130
67
S2-glass/E733FR Single Lap Joint specimen geometry
68
Typical evolution of maximum IR-TSA spatial gradients for increasing

fatigue cycles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
69
Typical evolution of maximum IR-TSA spatial gradients for increasing

fatigue cycles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
70
Area measures to evaluate the typical damage evolution as defined by

differential IR-TSA under increasing fatigue cycles . . . . . . . . . . . 136
71
Normalized effective bond length measures to evaluate the typical damage evolution as detected by IR-TSA under increasing fatigue cycles . 137
72
Fatigue damage indices of shear lap joints subjected to stress level of

0.65 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
xii
. . . . . . . . 132
73
Fatigue damage indices of shear lap joints subjected to stress level of

0.55 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
74
S-N curve progression for damage index of TSA area . . . . . . . . . 144
75
Constructed stochastic SN data based on two limited data sets and

previously calibrated MCM . . . . . . . . . . . . . . . . . . . . . . . 145
76
Proposed algorithm for generating stochastic S-N curves using previously calibrated MCMs. . . . . . . . . . . . . . . . . . . . . . . . . . 147
77
Schematic representation of the construction of stochastic SN curves.
78
Thermoelastic Stress Analysis for Fatigue of an IM7 Laminate . . . . 149
79
Thermoelastic Stress Analysis for Fatigue of an IM7 Single Lap Joint
80
Thermoelastic Stress Analysis for Fatigue of an IM7 single nested overlap joint . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
81
Thermoelastic Stress Analysis for Fatigue of a Unidirectional S2-Glass/Epoxy

Laminate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
82
Thermoelastic Stress Analysis for Fatigue of a Quasi-Isotropic S2Glass/Epoxy Laminate . . . . . . . . . . . . . . . . . . . . . . . . . . 150
83
Thermoelastic Stress Analysis for Fatigue of a Quasi-Isotropic S2Glass/Epoxy Single Nested Overlap Joint . . . . . . . . . . . . . . . . 151
84
Thermoelastic Stress Analysis for Fatigue of a Quasi-Isotropic S2Glass/Epoxy Single Nested Overlap Joint . . . . . . . . . . . . . . . . 151
xiii
148
149
SUMMARY
Many studies with Thermoelastic Stress Analysis (TSA) and Infrared Thermography, in Fiber Reinforced Polymeric materials (FRPs), are concerned with surface detection of hot spots in order to locate and infer damage. Such experimental
analyses usually yield qualitative relations where correlations between stress state
and damage severity cannot be obtained. This study introduces quantitative experimental methodologies for TSA and Digital Image Correlation to expand the use of
remote sensing technologies for static behavior, static damage initiation detection,
and fatigue damage in FRPs. Three major experimental studies are conducted and
coupled with nonlinear anisotropic material modeling: static and TSA of hybrid biocomposite material systems, a new stochastic model for fatigue damage of FRPs,
and fracture analysis for FRP single-lap joints. Experimental calibration techniques
are developed to validate the proposed macromechanical and micromechanical nonlinear anisotropic modeling frameworks under multi-axial states of stress. The High
Fidelity Generalized Method of Cells (HFGMC) is a sophisticated micromechanical
model developed for analysis of multi-phase composites with nonlinear elastic and
elastoplastic constituents is employed in this study to analyze hybrid bio-composites.
Macro-mechanical nonlinear anisotropic models and a linear orthotropic model for
fracture behavior using the Extended Finite Element method (XFEM) are also considered and compared with the HFGMC method. While micromechanical and FE
results provide helpful results for correlating with quasi-static behavior, analyzing
damage progression after damage initiation is not straightforward and involves severe
energy dissipation, especially with increasing damage progression. This is especially
true for fatigue damage evolution, such as that of composite joints as it is associated
xiv
with uncertainty and randomness. Towards that goal, stochastic Markov Chain fatigue damage models are used to predict cumulative damage with the new damage
indices proposed using full-field TSA image analysis algorithms developed for continuously acquired measurements during fatigue loading of S2-Glass/E733FR unidirectional single-lap joints. Static damage initiation is also investigated experimentally
with TSA in single-lap joints with thick adherends providing for new design limitations. Fracture behavior of a single-lap joint is also analyzed with different fracture
methods including the Virtual Crack Closure Technique (VCCT) and critical stress
techniques. The computational modeling, stochastic and experimental methods developed in this study have a wide range of applications for static, fracture and fatigue
damage of different FRP material and structural systems.
xv
CHAPTER I
INTRODUCTION
1.1
General
This chapter presents a literature review of significant studies on experimental, analytical and numerical studies characterizing the behavior of fiber reinforced polymeric
(FRP) material systems. In addition, quantitative experimental studies on the use
of Infrared Thermography and Thermoelastic stress analysis, micromehcanical and
macromechanical modeling approaches, static and fatigue damage detection in FRPs,
and stochastic modeling approaches for fatigue damage in FRPs. The final section of
this chapter will outline the objectives and research approach of the present study.
1.1.1
Fiber Reinforced Polymeric Materials
Fiber reinforced polymeric materials (FRP) consist of glass, graphite or carbon,

aramid, boron, silicon carbide, natural fibers, etc. held together in a structural unit
with a binder or matrix material and combined with fillers for mechanical performance or cost benefits. The binder or matrix material for FRPs is either a thermoset
or thermoplastic. Thermosets use polymer chains that cross-link and therefore do not
melt at high temperatures. Thermoplastics, on the other hand, use polymer chains
that do not cross-link allowing them to melt and harden again when cooled. Typical
thermosets are epoxies, polyester, or phenolic. Typical thermoplastics are polyimide,
polysulfone, polyetheretherketone (PEEK), or Polyphenylene sulfide.
Types of fiber reinforced composites include continuous, woven, chopped, and hybrid composites. Hybrid composites consist of mixed fiber types or fiber geometries.
Continuous composites consist of continuous fibers oriented in individual layers or
laminae and bonded together to form a laminate. Fabrication Processes for FRP
1
typically include open molds, autoclave, compression moulding, filament winding,

pultrusion, reinforced reaction injection molding, thermoplastic molding, resin transfer molding, and structural reaction injection molding. An autoclave is a heated
pressure vessel which takes a laminate through a temperature and pressure curing
program. Prepreg is typically used for autoclaved laminates. Prepreg is a tape of
precoated fibers with polymer resin. Achieving void conents of 0.1% is typical for
autoclaved laminates using prepreg. Pultrusion is a manufacturing process in which
unidirectional filaments are impregnated in resin and pulled through a heated die
to produce long prismatic structural components. Pultrusion and autoclaving were
used for fabricating the glass, carbon, and hybrid jute-glass fiber reinforced polymeric
materials used in this study.
1.1.2
Thermography
Thermography is the science of measuring temperature changes on the surface of

materials due to stress generated thermal fields (SGTF) or externally applied thermal fields (EATF). Thermography is a non-destructive investigation (NDI) tool that
allows remote sensing capabilities to detect imperfections or characterize materials.
This NDI tool typically requires a sensitive infrared camera capable of detecting
temperatures changes less than 50 mK. Sensitivities commonly reported during the
1980s were 0.1K, and in many cases trouble with data measurements was reported
due to ambient conditions. Since that time, methods to process and measure thermal
data such as Stress Pattern Analysis by Thermal Emission (SPATE) where points are
scanned in a point-by-point manner under adjustable computer control have developed into fast and accurate full-field Thermoelastic Stress Analysis (TSA) methods.
SPATE methods required 1-2 hours to obtain a single image scanning 50 points per
second. Thermoelasticity measurement systems today have a thermal resolution of
at least 1 mK for a full-field image with exposure times of 1 minute or less. This
temperature sensitivity allows for more sophisticated remote sensing capabilities to

exploit damage and damage progression in materials and structural components.
Materials must be excited to expose imperfections. Thermography can be used to
detect anomalies spatially because imperfections disrupt heat transfer. An excitation
source is the heat source that introduces energy to cause heat energy transfer to
occur. Excitation sources can be applied by externally applied thermal fields or by
mechanical means for stress generated thermal fields. In either case the thermal field
produced is dependent on the emissivity of the material investigated. Emissivity is
the ability and efficiency of a material to emit, reflect, or absorb energy. The choice of
EATF or SGTF depends on the application, service conditions, material properties,
and experience. In the field, the service conditions such as vibrations may be used to
excite the material. In many cases discrimination of damage effects is difficult, and
the particular method of IR-thermography used is based on the desire to illuminate
and excite the particular imperfections of interest.
1.2
Experimental Studies
1.2.1
Quantitative Thermography and Thermoelastic Stress Analysis in

Composites
Quantitative infrared thermography NDE methods are non-contact full-field techniques whereby an IR camera with digital sensors is used to detect small changes of
temperature due to different sources. The later can be in the form of an irreversible applied mechanical load, direct heat source, ultrasonic stress waves, among others. The
overall goal is to subject the structure or the material to thermomechanical deformation that produce spatial variations in the surface temperatures and allow correlation
between measured IR field and the stress or strain on the surface. Under adiabatic
and reversible conditions in isotropic materials, the application of a small cyclic load
will induce small and repeated variations in temperature that are proportional to the
sum of principle stresses. Kelvin [125] (Thomson, 1878) was the first to propose this
3
thermoelastic principal. He used thermodynamics-based derivations to obtain a linear

relation between the temperature change and the first stress-invariant. The recent
advances and affordability of charged-coupled-device (CCD) cameras, with fast acquisition systems, have lead to a powerful and quantitative thermoelastic stress analysis
(TSA) measurement techniques. The first invariant of the stress can be measured on
the surface of loaded coupons made of homogeneous materials. TSA has been also
applied in composite materials to measure stresses in thin laminated composites with
and without damage. Limited attention has been directed to quantitative TSA in
multi-layered fiber reinforced plastic (FRP) composite materials. In laminated composites, several difficulties may arise in the use of TSA, such as the effects of mean
stress and frequency dependent testing. Temperature diffusion and non-adiabatic
approaches have been proposed to model the thermoelastic effect in laminated composites. Potter et al. [104, 105] developed techniques to investigate the thermoelstic
effect in laminated composites. Bakis and Reifsneider [13] used laminate analysis and
micromechanical formulations to develop predictions of the thermoelastic signal, and
good agreement was found between their analytical analysis and experiments, with
differences due possibly to material heterogeneity and laminate surface conditions.
They also investigated the effect of the cyclic loading frequency on the thermoelastic
singal to find spurious non-adiabatic effects above 30 Hz. Dunn [44] used a mathematical model to account for the thermal conduction in the top epoxy layer of a
graphite/epoxy composite. Kyriakopoulos et al. [87] used heat conduction finite element analysis to quantify the TSA signal in the absence of adiabatic conditions. In
the case of thin laminates, the IR camera was found to detect temperature changes
on the surface ply alone. Van Hemelrijck et al. [68] utilized a non-adiabatic theory
approach that took into account the interlaminar heat transfer and obtained good
correlation between theoretical and experimental results of the surface temperature
for a cross-ply carbon/epoxy laminate. The technique was also used for qualitative
and limited quantitative studies. For example, Zhang et al. [140142] studied the
dependence of the thermal coefficients of thermal expansion, thickness of surface matrix resin, loading frequency, and changes in absolute temperature on the TSA signal.
They noticed an effect of the surface resin thickness. By altering the resin thickness,
they found that the TSA signal is highest without a surface layer and above a certain
thickness the TSA signal is constant-independent of specimen thickness. They related
this effect to a lack of heat transfer from the load carrying fibers to the surface at
greater thicknesses. They also researched stress concentrations in the knit-yarn fiber
intersections in a woven carbon/epoxy laminate. A mean stress effect was observed
in the glass/epoxy composite that was studied. Dulieu-Smith et al. [43] reported
test results of a full-field stress characterization on a woven tee-joint with FRP laminated fillets. Cunningham et al. [36] used a DeltaTherm IR detection system to
characterize damage around a circular hole in a laminated plate with unidirectional
E-glass/epoxy. Mackenzie [92] and Welch and Zickel [133] investigated the characteristics of the thermal radiation signal emitted from different surface coatings. The
solution of Mackenzie for the thermal wave problem characterized the IR flux amplitude from the surface as a function of IR reflection and thermal material properties of
the considered substrate-coat-air system. Using material properties of a typical paint
coating, a range of thicknesses and applied thermal frequencies was identified to allow
the coat to act as a strain witness layer. Barone and Patterson [17] proposed using a
polymeric coating to extract the strain field from TSA measurements. Their method
was applied for isotropic substrates. Good correlation between analytical solutions
and measured TSA responses was obtained for aluminum plates with circular holes.
El-Hajjar and Haj-Ali [48] proposed a technique to measure the sum of the direct
strains on the surface of thick section and orthotropic composites to the TSA signal
obtained from the surface of the specimen. Their method was verified experimentally
and compared favorably with finite-element (FE) simulations of notched and cracked
coupons. This method was used to verify damage studies in thick-section composite
materials, considered by Kilic and Haj-Ali [82] and Haj-Ali and El-Hajjar [49]. Ju
and Rowlands [78] applied the stress field solution expansion from Khalil et al. [71]
together with a thermoelastic stress analysis approach to determine the stress intensity factors (SIFs) of off-axis crack orthotropic composites. The thermoelastic stress
signal was assumed to relate to the in-plane direct stresses through two separate coefficients. They used finite element analysis to calibrate the two thermo-mechanical
coefficients to correlate the thermoelastic signals to the direct stresses. The SIFs
from Lekhnitskiis solution [88] via thermoelastic stress signals were calculated and
compared with those obtained from FE models. Tomlinson and Marsavina [127] and
Diaz et al. [39, 40] used Muskhelishvilis solution [98] to calculate the stress intensity factors (SIFs) in metallic materials using different IR-TSA techniques. They
provided methods to determine the location of the crack tip and studied the crack
closure and residual stress in their research. The SIFs were also determined from the
fatigue crack. Thick section pultruded composites were investigated and their TSA
technique showed accurate ability compared with analytical FE solutions. Haj-Ali
et al. [62] developed two IR-TSA methods to measure of the sum of surface strains
from the surface of FRP orthotropic composites. They showed that the mixed mode
stress intensity factors can be calculated using their method and the Lekhnitskiis
anisotropic elasticity solution for the stress field ahead of a crack tip in anisotropic
medium. Emery et al. [50] similarly developed calibration techniques relating the
in-plane strains and material properties to predict the thermoelastic signal in orthotropic laminates taking into account the surface characteristics. Dulieu-Barton
et al. [42] developed a temperature correction methodology for fiber reinforced polymeric materials and metals to account for a mean ambient temperature effect during
TSA measurements. Their methodology is better able to filter out effects outside of
the cyclic loading such as the operator handling the specimen during acquisition or
heat flow from the grips to the specimen during cyclic loading.
1.2.2
TSA for Damage in FRP Composites
Several studies have shown that infrared thermography is a powerful tool for evaluating damage in many applications with fiber reinforced polymeric materials. DulieuSmith et al. [43] evaluated a GRP tee joint using a combination of Thermoelastic
Stress Analysis (TSA) and FEA. The goal of the study was to correlate finite element to TSA without evaluating damage; however the method can be extended
to provide a tool for investigating damage initiation and propagation in composite
joints. Mackin and Roberts [93] tracked static damage progression in ceramic matrix
composites using TSA on double edge notched specimens. Bakis et al. [14] related
the residual strength, stiffness, and fatigue life to their corresponding damage states
obtained from photoelastic coating and thermal emission experiments for circular
notched graphite/epoxy laminates subjected to fully reversed fatigue loads. They
observed the damage initiated around the hole for quasi-isotropic and orthotropic
laminates. Matrix cracking and delamination patterns were different in both cases
due to the interaction between adjacent plies. Compared with photoelastic data,
the thermal emission was more sensitive to the minute deformations near fracture
paths in surface plies. Bremond et al. [31] also illustrated the advantages of infrared
theromography as a non-destructive method. Jones et al. [75] developed a damage
variable based on Thermoelastic Stress Analysis data to analyze damage in metal and
composite systems; this damage variable is showed excellent correlation to traditional
measurements of crack length in their study. Kageyama et al. [79] suggested a damage threshold approach based on 3D FEA, and used TSA with linear elastic fracture
mechanics to measure the crack propagation in notched carbon/epoxy laminates. Differential infrared thermography was proposed and used to track the damage in [45 ]
and [0/90 ] type graphite/epoxy laminates by Lohr et al. [91]. In their experiment,
the measured temperature was seen to decrease as the number of cycles increased
due to cracking in the epoxy surface layer. This effect was more pronounced under
higher frequencies and made quantitative TSA difficult for these composite material
systems. Adding a thick resin surface layer was found to stabilize the TSA signal and
attenuate the heat transfer from the carbon/epoxy inner layers. Cavaliere et al. [33]
applied TSA techniques to investigate the fatigue behavior of metal matrix composite
sheets joined by friction stir welding process. They stated that the infrared signal
relating the sum of principal stresses can be used to investigate stress concentration
and the stress intensity factors for notched metal matrix composite specimens. ElHajjar and Haj-Ali [48, 63] proposed a technique to measure the sum of the direct
strains on the surface of thick section and orthotropic composites to the TSA signal
obtained from the surface of the specimen. Their method was verified experimentally
and compared favorably with FE simulations of notched and cracked coupons. This
method was used to verify damage studies in thick-section composite materials, considered by Kilic and Haj-Ali [81, 82]. Johnson et al [73, 74] showed fatigue damage
progression in thick-section composites with Thermoelastic Stress Analysis (TSA).
TSA data was gathered to obtain a mean thermal effect that was related to fatigue
damage and damage evolution. The elastic modulus in pultruded materials was measured by an extensometer, and showed lower sensitivity under the fatigue loading
than TSA measurements. A limited number of specimens was examined; therefore, it
could not be used to predict fatigue damage evolution. This study was expanded on
by Wei et al. [132], and stochastic Markov Chain models were developed to characterize the fatigue damage in composite laminates from the cumulative IR-TSA data.
They proposed a method to predict the S-N curve and showed that TSA metrics
could be used to predict fatigue damage evolution.
1.2.3
Experimental Studies on Jute Hybrid Bio-Composites
Hybrid fiber reinforced plastic FRP systems are composite materials that can combine different forms of reinforcement layers, such as roving, continuous fillament mats
(CFM), woven fabrics, and braided preforms. The reinforcement combination is usually repeated through the thickness of the cross-section of a flat member. Fibers can
be made of carbon or glass or natural fiber yarns. Matrix materials are commonly
made of polyester or vinylester resin that includes additives such as glass microspheres
and clay particles. Pultruded composites can include thin and thick-walled members
(1/16 to 1 in.) having similar shapes to the standard steel beam shapes, such as
wide-angle, channels, and angle sections. Recently, FRP pultruded composite materials have been widely applied for civil and infrastructural engineering applications,
such as bridges, transmission towers, and structural components of buildings.
The demand for natural fiber composite systems is increasing, e.g., due to legislation
for partially decomposable vehicles to reduce landfills. Many car manufacturers are
using this green technology not only because it can be marketed as green, but also
because natural fiber systems can provide improved stiffness, weight, and insulation
at lower cost than glass fibers [24]. Frames, decking material, railings for parapet
wall, furniture sections, wall panels, roofing sheets, temporary shelters, post office
boxes are some examples of structural applications utilizing natural fibers [70]. Wool
et al. [120] suggests that monolithic roofs of natural fibers in a soybean resin can
resist hurricanes in addition to providing beneficial insulation properties. Different
combinations of fiber systems (Flax mats, Paper/chicken feathers, paper/corrugated
paper, paper/e-glass fiber) for composite beams were investigated by Dweib et al. [46]
as manufactured by the Vacuum Resin Assistant Transfer Moulding (VARTM) process
to show the potential for different structural applications with natural fiber systems.
Van de Velde et al. [128] developed pultruded systems from flax fibers.
Durability of natural fiber composites is a concern for any exposed application.
9
Several studies [21, 97] have investigated the durability of composites in an alkaline
environment, three years natural and 1000h accelerated UV, simulated sunlight, immersion at 70C in water for 30 days, and freeze and thaw cycles (-25 to 20C), etc.
These studies show that the severity of aging is most detrimental in accelerated water
tests. Also in some natural fiber systems fungal infestation can occur at cut edges
of weathered composites. Failure of natural fiber composites due to delamination or
fibre swelling in wet conditions has spurred further work into chemical treatments
to increase moisture resistance (increased performance in different aqueous environments) in the natural fibre systems [9, 22, 35, 47, 129, 138]. Sabeel Ahmed et al. [8]
investigated various combinations of jute-glass laminates evaluating tensile, flexural,
and interlaminar shear properties as measures of performance. The effect of hybridizing was investigated to minimize overall cost of the laminates taking advantage
of the lower cost of jute while greatly increasing resistance to moisture absorption.
They found that various jute-glass hybrids out perform a homogenous glass or jute
reinforced polymer systems when cost and moisture resistance is taken into account.
Jute fibers are popular as reinforcement materials because of their increased elastic
modulus and higher elongation at fracture as compared to other natural fiber systems. Many sources list the longitudinal Youngs Modulus of Jute Fibers from 20 40 GPa; discrepancies in the literature are possibly due to assumptions of a circular
cross-section [8, 23, 107, 113, 118] in addition to the inherent mechanical variability
of natural systems. The modulus is significantly lower than the modulus of E-Glass
fibers which have a longitudinal modulus of approximately 73GPa.
1.2.4
Damage in FRP Lap Joints
Recent innovations in advanced aerospace structures can be attributed to the manufacturing of composite materials with superior strength-to-weight ratios. Engineering
10
aircraft structures with composite materials requires a detailed knowledge of durability and damage tolerance of individual structural components and especially fiber
reinforced polymer (FRP) joints. Traditional mechanical testing methods using extensometers and strain gauges of composite joints may only measure linear loaddeformation responses to failure giving no indication of overload or failure initiation.
Non destructive evaluation tools offer a significant refinement over traditional mechanical tests, such that failure initiation of critical components can be detected and
identified with some limitations. This is especially important since the bonding condition cannot be easily checked.
Single lap joint geometry has been traditionally used as a control to justify changes
in design towards improving static and fatigue performance. In many cases FEA is
used to investigate the stress distribution differences within the bond for alternative
joint geometries to better understand the effect of joint geometry on joint performance. For example, Zeng et al. [139] developed a wavy composite lap joint as an
alternative to traditional lap joints or adhesive joint geometries with tapered edges to
avoid the load eccentricity and the associated singular peel stresses at the joint ends.
The wavy lap joint resulted in compressive peel stresses at the joint ends altering the
failure progression so that there was no indication of damage initiation before final
failure. Comparing this to the traditional lap joint, crack initiation was noticed from
the load/displacement relationship and visual inspection of cracks. Avila et al [12]
used FEA to make correlations between stress distributions in wavy lap joints and
single lap joints with the failure loads in E-glass/epoxy composites. They noticed
a 41% higher load carrying capacity for wavy lap joints over conventional single-lap
joints attributed to a more uniform stress field with compressive peel stresses in the
wavy lap joint. Fessel et. al [51] showed significant improvements in overall joint
strength for the reverse-bent joint over the traditional lap shear joint for several steel
alloy substrates with different overlap lengths. They used FEA to evaluate stress
11
distributions within the bond and discussed possibilities for improved joint strength
by modifying joint geometries to achieve more uniform stress distributions instead
of high localized stresses at the joint ends with a relatively unstressed central region. Borsellino et al. [30] showed for a given resin longer cure times correspond to
a significantly stiffer, stronger resin as shown by stress-strain relations. The authors
also showed some evidence of capturing stabilization by investigating changes in failure (adhesive/cohesive) mechanisms with extended curing times as viewed by surface
inspections of failed single lap joints. The FE was used to evaluate internal stress
distributions, and experimental evaluation was based on mechanical testing (flexural
modulus, ultimate failure stresses, or impact resistance). Da Silva et al. [37] compared basic double lap geometries with an inside tape and adhesive fillet design with
various resins using experimentally determined failure loads and FEA to evaluate the
internal stress distributions due to combined temperature and mechanical loads using titanium and carbon fiber composites. Many experimental methods were used to
evaluate overall joint performance based on more traditional mechanical testing and
ultimate failure states.
1.3
Analytical and Numerical Studies
1.3.1
Nonlinear Anisotropic Micro-macromechanical Modeling
Many experimental and analytical studies have been focused on the nonlinear response of laminated composite materials. Macro-level theories are formulated to
characterize the nonlinear response by idealizing the composite as an anisotropic homogeneous medium. Petit [102] proposed an incremental approach for the analysis of
symmetric laminates under uniform membrane loading. In their approach, the elastic
constants of a lamina are determined as a function of the lamina strains. Lamina
failure is identified when any of its strain components exceeds the experimentally
obtained limiting strain value. Once a failure mode is detected in the lamina, the
12
corresponding tangent stiffness is set to a high negative value in order to achieve

stress unloading in this direction. Their analyses predicted the overall shape of the
nonlinear response quite well, while the laminate ultimate failure loads were predicted
with acceptable accuracy. Hahn and Tsai [55] used the complementary energy density polynomial function, for a lamina under a plane-stress state, to derive a nonlinear
stress-strain relation for laminated composites. An additional fourth-order term of
the axial-shear stress is added to the polynomial function. This term represents the
nonlinear shear strain. Interaction terms between the three stress components were
neglected. Hahn [54] extended this work to model the nonlinear behavior of laminates, and considered the effect of material nonlinearity on the buckling load of a
symmetric laminate. Hashin et al. [67] formulated a new nonlinear constitutive model
where inelastic transverse and axial shear strains exist in a lamina under plane stress
conditions. Each inelastic strain component is an independent quadratic function in
terms of the stress invariants raised to a general power. Ramberg-Osgood representation of nonlinear stress-strain curves was used to express the nonlinear axial-shear
and transverse stress-strain relations. Predicted nonlinear behavior compared well
with experimental results for different laminated composite materials. Jones and
Nelson [77] proposed a material model based on strain energy that accounts for nonlinear behavior under biaxial states of stress. This model can only be applied to
multi-axial stress states where the strain energy is lower than the maximum strain
energy generated in the uni-axial test results. Since this is often the case when the
composite is subject to general multi-axial stress states, different approaches were proposed to overcome this difficulty, Jones and Morgan [76], Abu-Farsakh [7]. Amijima
and Adachi [10] introduced a simple technique to represent the shear nonlinearity
using piecewise linear segments. Sandhu [112] introduced a technique for solution
of the nonlinear laminate equations using piecewise cubic spline functions to represent the lamina uni-axial test data. He also proposed a predictor-corrector iterative
13
method. Nahas [99] presented a similar predictor-corrector technique in which the

basic uni-axial stress-strain relations are allowed to have a general nonlinear representation. Kuppuswamy et al. [86] used the Richard and Abbott [106] representation
for the nonlinear uni-axial stress-strain relations in a three-dimensional finite element
analysis. However, their study uses uncoupled one-dimensional stress-strain relations
which do not reflect interaction effects. Pindera and Herakovich [103] and Mathison
et al. [96] derived a nonlinear plane stress constitutive model using orthotropic endochronic theory. This theory is based on irreversible thermodynamics with internal
variables. Dvorak and Rao [45] proposed a plasticity theory for fibrous composites
under axisymmetric deformation. Plastic dilatation and deformation of the composite in the fiber direction were accounted for. A yield function was formed using the
stress invariants of a transversely isotropic material and simple hardening and flow
rules were derived. Griffin [53] extended Hills anisotropic plasticity theory, and used
the associated plasticity flow rule to determine the plastic increment of strain. A
Ramberg-Osgood uni-axial stress-strain relation was used to model nonlinear hardening. Sun and Chen [119] developed a one parameter orthotropic plasticity model for
thermoplastic and metalmatrix composite laminates in plane stress. In this model,
one-parameter plastic-potential scalar function is proposed. Only transverse and axial shear stresses are involved in the plastic potential; therefore, plastic deformation
in the axial direction of the lamina is neglected. In addition, a power law relation
between effective plastic strain and the effective stress is proposed; as a result, a total
of three material parameters are needed to fully describe the lamina plastic behavior.
A nine-cell cohesive micromodel is developed similar to the doubly periodic multiphase composite media fully described and reviewed by Aboudi [3, 4, 6], and is a simplified version of the more general case of the High Fidelity Generalized Method of
Cells given by Haj-Ali [56]. The High Fidelity Generalized Method of Cells (HFGMC)
is a computational nonlinear micromechanical model for multi-phase composites. A
14
more thorough review review of the HFGMC model can be found in Aboudi [3] and in
Haj-Ali and Aboudi [58] where the modified form now includes nonlinear algorithms
with tailored stress correction schemes, iterative micromechanical solution methods
for rapid error reduction, and multi-scale features extending the capabilities of the
model for greater efficiency in capturing the local and global load re-distribution that
is associated with nonlinear and damage effects in large-scale structural computational finite element models consisting of doubly periodic multiphase composite media. Classical nonlinear micromechanical models of multi-phase materials are capable
of capturing the nonlinear effective response of the material but neglect the local spatial variation of the deformation fields within the representative microstructure. The
HFGMC has the capability of predicting local nonlinear mechanical behavior of periodic multi-phase materials subject to remote loading through a modeling framework
that allows for a detailed description of the constituents and spatial deformations between the phases. Multiphase composites are analyzed within the HFGMC framework
with nonlinear elastic, elastoplastic and viscoplastic constituents Aboudi2004.
The HFGMC has been used to analyze fully coupled electro-magneto-thermoelastic multi-phase composites [5], fiber-matrix debonding in metal matrix composites [19], thermomechanical analysis of internally cooled structures at elevated temperatures for aerospace engine applications [11], adhesively bonded composite joints [20],
and internal pore pressure for the analysis of foam insulation of the external tank of
the space shuttle [18].
Experimental and analytical studies have been performed to characterize the effective material properties and behavior of pultruded FRP composites. Herakovich
and Mirzadeh [69] studied the effects of the spatial distribution of fibers and fiber
volume content on the effective properties of pultruded graphite/epoxy composites.
They observed that the fiber and resin rich area were distributed non-uniformly. In
15
addition they found fiber waviness, which reduces the stiffness and strength of the pultruded composites. Bank [15,16] used the Iosipescu shear test to characterize in-plane
shear modulus and strength of glass fiber pultruded specimens having vinylester and
polyester matrix. The specimens were cut from the webs and angles of wide-angle
pultruded beams. Pronounced nonlinear stress-strain relations were shown. Wang
and Zureick [130] characterized the tensile behavior of coupon specimens cut from
different locations of a pultruded I-shaped beam. Flaws such as high void content
and uneven distribution of reinforcement was significant in this material system. The
specimen size effects on the effective properties were also examined. Tomblin and Barbero [126] performed analytical model to predict unidirectional compression strength
of pultruded composites. Nonlinear responses were shown mainly due to fiber microbuckling. Nonlinear stressstrain behavior under multi-axial static loading can
occur in pultruded FRP materials and structures. Haj-Ali and Kilic [59] conducted
extensive off-axis tests finding a pronounced nonlinear material response evident in
all off-axis angles of thick-section pultruded composites including the uniaxial specimen (0 ). This material nonlinearity was attributed to the low fiber volume fraction
(FVF) and behavior of the soft polymeric matrix.
1.3.2
Analytical and Computational Modeling of FRP Lap Joints
Hart-Smith [65] examined the effect of plasticity in the adhesive in FRP lap joints,
an imbalance of the stiffness of the adherends, bond length, fiber stacking sequences,
adherend thermal mismatches, and other effects in his analytical parametric study.
Hart-Smith notes three distinct and characteristic failure modes in single lap joints:
adherend failure due to in-plane stresses from the eccentricity in the load path, failure
of the adhesive in shear, and failure of the adhesive due to peel stresses. Elastic studies
have indicated that joint failure emanates from shear failure of the adhesive; however,
the inclusion of adhesive plasticity in the analysis indicates that failure is associated
16
with adhesive peel stresses. If the interlaminar tension strength is lower than the
peel strength; however adherend failure is expected. Hart-Smith [66] expanded this
study to double-lap, single-lap, scarf and stepped-lap configurations under tensile,
compressive, and in-plane shear load conditions covering joint parameters such as:
adhesive plasticity (elastic-plastic shear stress model), adherend stiffness imbalance
and adherend thermal mismatch. Again, he focused on failure modes such as the case
when the adherend fails outside the joint (1), the ultimate shear strain of adhesive
is exceeded resulting in bond failure (2), and peel stresses induced at ends of joint
exceed the interlaminar tension strength of the laminate resulting in splitting failure
of the adherend (3). The main modification suggested to alleviate these peel and shear
stress intensity and failure mode is to taper the end of the outer adherend. HartSmith suggests that peel stresses are only a problem if joint efficiency is low; however,
peel stresses are still important in more efficient joint types especially for fracture
analysis. Dickson et. al [41] also developed a parametric study using closed form
analyses, finite element analysis, and photoelasticity to verify their solutions. They
included the effects of stress through the thickness of the adherends but neglected the
effects of adhesive plasticity. Instead they chose to include a plastic zone approach in
their analytical model which ignored the presence of peel stresses in the adhesive and
assumed that shear stresses are constant in the zone as an approximation. Again,
Dickson et. al focused on ultimate joint strength instead of failure initiation in this
study. Lai et al developed a closed form solution to obtain fracture parameters for
investigating debonding in the cracked lap shear joint problem. Their analytical
method is compared to finite element results and fatigue experiments. The solution
is similar to Goland and Reissner approach [52] for a classical single lap shear joint
problem involving only linear behavior of the adhesive, but adds fracture parameters
for investigating crack growth. Campilho et al. [32] investigated joint efficiency using
a parametric finite element study on internal stress distributions for different overlap
17
lengths, plate thickness, and stacking sequences of single lap joints similar to HartSmith. The experimental part of their study focused mostly on ultimate failure
stresses of the joints instead of failure initiation.
Although strength based approaches for failure in single lap joints [65, 66] has
shown excellent correlation with experimental results, fracture mechanics approaches
for fiber reinforced polymeric composites have gained more popularity in the past
two decades [85]. Linear elastic fracture mechanics (LEFM) emerged from linear
elasticity to solve for the stress and deformation fields around cracks in continuous
media. These solutions and fracture-based material properties are used in design
problems of materials and structures with defects. By using Westergaards elasticity
solution [134], Irwin [72] shows that the stress field expansion near the crack tip is
related to a constant multiplied by the coefficient of the square root of the distance
from crack tip. This constant is called the stress intensity factor (SIF), K. The SIF is
shown to be directly related to the strain energy release rate. Sih, Paris and Irwin [115]
employ Lekhnitskiis [88] solution to relate strain energy release rate to the SIFs for
anisotropic materials. The mixed-mode case in absence of the tearing mode is also
presented. They show that for anisotropic materials where the crack is aligned in one
of the material symmetries, the SIF in mode-I and II are not cou-pled. In the case
where the crack is not aligned with material symmetry, the SIFs are both coupled and
need to be solved simultaneously. Dally and Sanford [38] propose an experimental
method to determine the mode-I SIF in isotropic materials using a strain gauge. The
Westergaard stress function [134] is used to calculate the mode-I SIF based on strain
gauge measurements. To obtain an accurate SIF, they determine the optimal location
and orientation of the strain gauge needed to elimi-nate the effect of non-dominant
terms and increase the contribution of the mode-I SIF. Shukla et al. [114] extend
Dally and Sanfords method to orthotropic materials. They use Airys function for a
two-dimensional orthotropic body and solve for the strain field equations. Rybicki
18
and Kanninen [111] develop a virtual crack closure technique (VCCT) to evaluate
the SIFs. The technique is based on the crack closure integral, which can be used
in a finite element analysis (FEA) with a coarse mesh. Damage onset and growth in
composite components at material or geometric discontinuities is related to Mode I, II,
and III delaminations. Different mixed mode fracture criterion has been developed for
crack growth due to a combination of Mode I and Mode II delaminations. The Virtual
Crack Closure Technique (VCCT) criterion uses the principles of linear elastic fracture
mechanics (LEFM) and, therefore, is appropriate for problems in which brittle crack
propagation occurs along predefined surfaces. VCCT is based on the assumption that
the strain energy released when a crack is extended by a certain amount is the same
as the energy required to close the crack by the same amount. Stroud et. al [84]
used fracture mechanics and probabilistic analyses to investigate the strength of a
single lap joint exploring the effects of configurational and material uncertainties on
the strength of a single lap shear joint.
1.3.3
Fatigue Damage Models for FRP Composites
Spatial damage distributions during fatigue experiments of uniform samples are usually scattered in part due to the material heterogeneity. A probabilistic damage
approach is therefore advantageous over a deterministic method in order to account
for the relatively wide variability of test results. Bogdanoff et al. [2629] were the
first to introduce a new cumulative damage model using stochastic processes. A
Markov chain model was used to take into account the variability (e.g., manufacturing and service loads). They further pointed out that the life testing in which
time to failure was recorded does not provide enough information for the member
function (mf) behavior. Thus, they divided the probability transition matrix into
different segment numbers to improve accuracy. Results of different probabilities in
the transition matrix of the Markov chain model were investigated analytically. They
19
compared previous experimental work done by other researchers with Markov chain
models. Bogdanoff and Kozin [28] called the cumulative damage model the B-model
and converted it to a nonstationary model using a time transformation condensation
method. Polynomial equations were used for mapping the time in different domains.
To maintain the definition of the transition matrix, a condensation method was used
in their models. This allows the stationary Markov chain to become a nonstationary
model. Kozin and Bogdanoff [83] employed the B-model together with the Paris law
to describe the fatigue process. Rowatt and Spanos [110] applied a cumulative damage model to laminated composite materials. The time transformation condensation
method was also employed in their work. They investigated the evolution of compliance in composite materials subjected to a constant amplitude fatigue loading. The
change in compliance was related to the individual damage mechanisms of a critical
element within the laminate to achieve a life prediction of the material system. Their
experimental data was based on a single mid-specimen axial extensometer measured
in a straight composite coupon. Xi et al. [108, 137] used a cumulative damage model
together with the R-curve to predict the damage evolution of a three point bend
concrete beam. They used a geometric series to generalize transition probabilities in
Bogdanoff and Kozins approach. A recursive expression was proposed to obtain the
probabilities in the transition matrix. The R-curve was obtained in order to provide
the crack extension, which was required in the recursion, for a specific specimen geometry and given loading conditions. Wu and Ni [136] studied three different damage
approaches: Markov chain model, Yangs power law model, and a polynomial model
for fatigue crack growth. They performed constant and random amplitude loadings
on aluminum alloy compact tension C(T) fracture specimens, and they used the three
models to predict the fatigue behavior of the specimens. For the case of random loading, the polynomial model fitted the experiments better than the Yangs model. They
pointed out that the nonstationary Markov chain model was sufficient to describe the
20
fatigue process and also suitable for random loading. Tanimoto [124] applied different
stress levels to carbon/epoxy laminates and observed that the fatigue life had more
scatter in the 105-106 cycle range. A Weibull probability model was used to construct the probability of failure at different stress amplitudes through the fatigue life.
The dominate failure mode was delamination. Liu and Mahadevan [94] utilized Monte
Carlo simulations combined with Miners rule that took into account the coefficients in
Tsai-Hill failure criterion in order to estimate the fatigue life of composite laminates.
They assumed the random variable such as ply thickness, elastic modulus, and ply
angle as normal distribution for Monte Carlo simulations. Their numerical method
had good agreement with the experiments conduced by Mandell and Samborsky [95].
Liu and Mahadevan [89] employed a nonlinear fatigue damage accumulation method
with a stochastic S-N curve technique to predict the fatigue life of metallic materials
under various stress levels. The fatigue data was obtained from the literature for
different materials with variable or constant amplitude loadings. A required coefficient for formulating the nonlinear damage accumulation rule was calibrated with
the experimental fatigue data. The coefficient accounted for material properties and
applied stress levels, which were not usually considered in a typical linear damage
accumulation rule. The Karhunen-Loeve expansion method [90] was conducted to
generate their stochastic S-N curves. Numerical simulations using the proposed damage method yielded satisfactory agreement with the experimental fatigue life. The
majority of proposed cumulative damage models used to evaluate material behavior
under fatigue are mainly based on traditional contact based measurements, including
strain gauges and extensometers. Rowatt and Spanos [110], for example, used extensometer data gathered from 17 coupons to investigate changes in compliance that
was related to the stochastic evolution of fatigue damage in composite materials. Wu
and Ni [136] used dial gages to measure fatigue crack growth in metallic C(T) specimens. Traditional contact based may not be practical for all specimen geometries, and
21
non-traditional tools may also be more sensitive to damage especially in composite

joints. Tamuzs et al. [122, 123] conducted fatigue tests on woven orthotropic composite laminates in off-axis loadings. They developed a quadratic master curve function
to describe the strength, deformation, and degradation of the stiffness. This master
curve was obtained by the best fit of the experimental data through the fatigue life
of the laminates, and showed that the reduction of the stiffness before fatigue failure
does not depend on the loading angles with respect to the material angles. An S-N
master curve was also developed by replacing the applied stresses in the conventional
S-N curve with a modified stress function. The S-N master curve was an innovative
alternative way of representing damage accumulation using a stochastic approach.
However, the variability in composite fatigue data was not fully accounted for.
1.4
Objective
The objective of this manuscript is to develop new experimental and computational

methods for analyzing the behavior of FRP composites. For the experimental part
of this study, quantitative methods using Thermoelastic Stress Analysis and other
experimental methods are developed for validating new computational modeling approaches for a new class of materials. Jute hybrid bio-composites were chosen for
the experimental and analytical part of this study. Nonlinear anisotropic micro and
macromechanical models were developed to capture the multi-axial response of the
material. The last part of this manuscript is focused on new quantitative methods for
analyzing lap joints using TSA. First, lap joints are tested statically, and changes are
correlated with real damage effects using photomicroscopy. Then, stochastic methods
are developed with new methods for processing in-situ TSA cumulative fatigue data.
22
1.5
Present Study
Chapter 2 of the present study develops a quantitative thermoelastic strain analysis method for composite materials. El-Hajjar and Haj-Ali verified fracture mechanics computational models for thick-section pultruded E-glass/polyester and Eglass/vinylester composites with experiments using this quantitative strain analysis
method. In chapter 2 this study provides this thermomechanical calibrations for
prepreg S2glass/epoxy, Carbon/epoxy, and pultrueded E-glass/polyester. This work
is then extended to a new class of materials investigating the linear and nonlinear
mechanical behavior of Jute hybrid bio-composites, and new nonlinear micromechanical and macromechanical modeling approaches are developed for the analysis of the
materials. Two of these nonlinear anisotropic homogeneous models assume that the
material is under a state of plane stress. The third is a nonlinear 3D micromechanical
model using a Nine-Cell micromechanical model, a special case developed by HajAli [56] of the more general High Fidelity Generalized Method of Cells (HFGMC).
The HFGMC micromechanical modeling framework was proposed by Aboudi [101]
and later modified by Haj-Ali and Aboudi [58] to include nonlinear algorithms with
tailored stress correction schemes, iterative micromechanical solution methods for
rapid error reduction, and multi-scale features extending the capabilities of the model
for greater efficiency in capturing the local and global load re-distribution that is associated with nonlinear and damage effects in large-scale structural computational
finite element models consisting of doubly periodic multiphase composite media. The
models from are calibrated from uniaxial experiments and validated in the third part
of this study with Thermoelastic Stress Analysis (TSA) and Digitial Image Correlation (DIC) for a full-field stress-strain analysis of plates with a hole. Tests are
conducted to generate the stress-strain nonlinear response up to ultimate failure and
three nonlinear constitutive material models are used to predict the multi-axial stressstrain behavior of bio-composites. Chapters 4 and 5 cover damage in FRP single lap
23
joints using TSA. Future damage detection techniques in aging aircraft will require
quantitative and noncontact nondestructive evaluation (NDE) methods especially for
composite components. Infrared (IR) thermograpy techniques have been qualitatively
used to assess and indirectly infer the durability of structural systems. IR-NDE tests
to ascertain the health and integrity of the structure over time, especially around connection areas and other critical locations where potential damage can occur. Quantitative NDE-IR methods have received limited attention, especially in composite
materials, because these require rapid data acquisitions and signal processing coupled with high resolution and full-field IR. These are past limitations of electronic
hardware and the software. One objective of this research was to examine composite components under loading and extract deformation measures from the emitted
IR field of the tested components. A research collarobation with Lockheed Martin
for non-destructive evaluation of composite lap shear joints led to a development of
thermoelastic stress analysis techniques for evaluation aerospace structures. Chapter
4 presents a study on infrared thermography for failure initiation and progression
in composite lap shear joints. This study also presents some generic finite element
results for analyzing bond characteristics in lap joints.
Chapter 5 expands on the previous chapter by implementing TSA on FRP single
lap joints for fatigue damage. IR-TSA techniques are employed to capture cumulative
fatigue data and a damage index is defined for the quasi-isotropic S2-glass/E733FR
composite lap-joints subjected to constant amplitude fatigue loading. Markov chain
models are employed with the TSA damage index to predict fatigue behavior for a
maximum stress corresponding to 65% and 55% of the ultimate stress. Next, new
stochastic S-N curves utilizing the calibrated Markov damage model are also introduced. The TSA and mechanical testing setup and test procedures are first discussed. The proposed IR-TSA damage index is defined based on the area reduction
of stresses. The numerical results from the experiments are then utilized along with
24
Markov chain theories to predict the fatigue behavior in S2-glass/E733FR lap-joints.

Chapter 5 presents infrared thermography for fatigue damage detection in FRP composites with stochastic methods for analyzing this fullfield data. Chapter 6 presents
the major outcome of this study and future work.
25
CHAPTER II
THERMOELASTIC STRESS ANALYSIS FOR

QUANTITATIVE ANALYSIS
2.1
Methodology
A experimental thermoelastic stress/strain analysis (TSA) technique is presented to

measure normal stresses or the sum of the normal surface strains in FRP composites.
The method proposed by El-Hajjar and Haj-Ali [48] can be used to verify computational models by providing a full-field measurement of the strain (or stress) invariant.
This chapter will present the derivation for this quantitative TSA method, steps to
acquire a thermomechanical calibration constant, and a verification of a computational models under a multi-axial state of stress for different material systems under
different loading frequencies. This work is presented as a continuation of the work
from the Masters Thesis by Shane Johnson [73], and this experimental work is presented here serving as a basis for utilizing TSA as a quantitative method for verifying
computational models under multi-axial states of stress. All the experimental work
in this section is was newly acquired to verify the accuracy of the method with different material systems and at different frequencies and stress or strain ranges. In
addition to this method, the same experimental calibration is executed without the
surface coating to show that two calibration constants would then be required for
an orthotropic composite. Derivation is also introduced for the case of an uncoated
multi-layered composite. This derivation was used with Lekhnitskiis elasticity solution to quantify the full strain field and determine mixed-mode stress intensity factors
(SIFs) for crack tips in composite plates subjected to off-axis loading in the fracture
study by Haj-Ali et al [62].
26
In general, to obtain the thermomechanical calibration constant the composite material tested is given an isotropic surface coat that experiences the same in-plane strain
conditions as the individual lamina in a perfectly bonded laminate or pultruded composite. During the calibration, the infrared (IR) signal from the coat is measured, and
strains in the axial and transverse loading directions are recorded while the specimen
under a uniform state of stress is cyclically loaded. This quantitative thermoelastic
strain analysis method allows a single material constant, comprised of the thermomechanical properties of the composite, to be calibrated such that the measured IR
signal can be related to the sum of the in-plane strains. The theoretical derivation of
this method developed by El-Hajjar and Haj-Ali [48]assumes steady state, adiabatic
conditions and is applied to the case of multi-layered composite materials. The proposed method is validated by its application to Carbon/epoxy, S2-glass/epoxy, and
thick-section E-glass/polyester composites under multi-axial states of stress.
2.2
Derivation
The proposed formulation directly relates the TSA signal to the sum of the direct
strains in a coated, layered composite medium. A perfect bond is required for the
validity of the method because the at each material point the same in-plane strains
must be experienced by the surface coating and individual laminae, and debonded
sections of a composite woud cause non-adiabatic effects like friction and rubbing.
The thermomechanical material response of the isotropic surface coating is considered linear elastic and independent of the mean applied stress and loading frequency
where the axial and transverse coefficients of thermal expansion are equivalent. Figure 1 schematically illustrates a typical layered composite with surface coatings and
different orthotropic layers. If the elastic stiffness of the coat is small compared to
the overall stiffness, the measured mechanical response is primarily due to the substrate. While assuming that the surface coating is responsible for the thermoelastic
27
response (1 = 2 ), the TSA signal is interpreted to represent the change in the sum
of the direct in-plane strains of each layer. The new material constant that is derived
relates the TSA signal to the change of the in-plane strains and can be calibrated
experimentally from TSA measurementson speciments under a uniform state of stress
when the in-plane strains are known. In-plane strains are easily measured during the
calibration with a bi-axial strain gauge mounted on the back surface of the same
specimen under a uniform state of stress.
Figure 1: Schematic cross-sectional view of a layered orthotropic composite with

surface coatings
Displacement continuity is assumed between the individual laminae to account
for the same in-plane strains in all the layers when a uniform in-plane loading is applied during the thermomechanical calibration. Assuming traction continuity results
in equivalent uniform out-of-plane stresses with the composite and surface coating
layer. The 3D lamination theory developed by Pagano [100], is used in this formulation to enforce traction and displacement continuity between indivual laminae while
accounting for equivalent linear anisotropic stress-strain relations for the substrate.
A basic equation governing the thermoelastic effect is given by Wong et al. [135]:
28
Qi,i = T
ij
ij o C T + o R
T
(1)
Where Qi is the heat flux through the surface whose outward directed normal is
ni , T is temperature, ij is the stress tensor, ij is the strain tensor, o is the density,
C is the specific heat at constant deformation, and R is the heat production rate per
unit mass of the internal heat sources. The thermomechanical constitutive law for a
linear anisotropic material is given by Hookes law:
dij = Cijkl dkl Cijmn mn dT
(2)
Where, Cijkl is the fourth rank tensor of the material constants, mn are the
coefficients of thermal expansion and dT is the change in temperature. If the material
properties are assumed to remain constant with change in temperature, then the stress
change in temperature is simply:
ij
= Cijmn mn
T
(3)
Assuming adiabatic conditions and the absence of internal heat sources, the heat
flux through the surface Qi,i = 0, and the specific heat at constant deformation R = 0,
Equations 3 and 1 are combined to give:
o C
T
T
= Cijmn mn ij
(4)
Next, assuming small strain theory, and using the incremental form of Equation
2 to express for Cijmn dij , Equation 4 is rewritten as:
o C
dT
T
= mn [dmn + Cmnpq pq dT ]
29
(5)
For the special case of an in-plane transversely isotropic coat under a state of
plane stress, the in-plane material properties are equal, 11 = 22 = , C11pq =
C22pq , d33 = 0. Placing theses conditions on Equation 5, it is then seen that the
thermoelastic effect is dependent only on the in-plane properties:
o C
+ Cmnpq mn pq ]dT = (11d11 + 22 d22) = (d11 + d22 )
T
(6)
The general, the stress-strain relation for the in-plane isotropic surface coating is:
11

22

33
=

23

13

12
C11 C12 C13
C12 C11 C13
C13 C13 C33
C44
C44
(C11 C12)/2
11
22
33
23
13
12
(7)
The plane stress assumption in this layer enables expressing the out-of-plane strain
33 in terms of the in-plane strain components:
33 =
C13
C13
(11 + 22) =
; = 1, 2
C33
C33
(8)
Therefore, the first stress invariant can be expresed as:
= = 11 + 22 =
2
(C11 C33 + C12C33 2C13
)
(11 + 22)
C33
(9)
From an experimental perspective, the infrared detector measures an un-calibrated

TSA signal, S, that is linearly related to the radiant photon flux emitted due to a
surface temperature change (S
T
T
). Equation 6 can be used to relate the

30
TSA signal to the incremental change of the first stress invariant through a parameter
k :
= k S
(10)
Substituting Equation (9) into (10) results in:
2
(C11C33 + C12 C33 2C13
)
(11 + 22) = k S
C33
(11)
Next, the above elastic constants of the surface layer are used to relate the TSA
signal to the direct in-plane strains instead of the stress invariant. This is important
since all orthotropic layers have the same in-plane strains in a medium subjected to
in-plane loading. A new constant k can then be defined as:
k =
C33
k
2
(C11 C33 + C12C33 2C13
)
(12)
This enables formulating a new thermoelastic equation, similar to Equation 10,

in terms of the in-plane strains using the new relationship:
= k S; = 1, 2
(13)
For a composite material with an orthotropic top surface and no surface coating,
Eq. (6) cannot be used (i.e., no scalar relation exists between the sum of stresses
and the temperature gradients for the orthotropic case). Two separate material coefficients are now needed. Therefore, Eq. (6) can be rewritten for an orthotropic case
in the form:
dT = A1 11 + A2 d22
31
(14)
In order to calibrate the above two constants, the stresses can be replaced by
strains.
dT = A1 (C1111 + C12 22 + C1333) + A2 d(C12 11 + C2222 + C23 33)
(15)
where the Cij is the material orthotropic stiffness. Imposing a state of plane stress,
the stress-strain relations of an orthotropic material can be expressed as:
11
22
12
Q11 Q12
=
Q12 Q11
0
0
11

0
22

12
Q66
The out-of-plane strain 33 can be expressed as:
33
(16)

13
23
13
23
13
23
=
11
22 =
Q11 +
Q12 11
Q12 +
Q22 22(17)
E11
E22
E11
E22
E11
E22
Therefore,

13 + 12 23
13 21 + 23
33 = B1 11 + B2 22; B1 =
; B2 =
1 1221
1 1221
(18)
Substituting Eq. 15 into Eq. 18 yields:
dT S = H1 d11 + H2 d22
(19)
H1 = (A1 C11 + A2 C12 + A1 B1 C13 + A2 B1 C23)
(20)
H2 = (A1 C12 + A2 C22 + A1 B2 C13 + A2 B2 C23)
(21)
Equation (19) implies that one can calibrate the coefficients, H1 and H2 , separately
and apply superposition, assuming that application of the model remains in the linear
range of the material.
32
In the axial direction, Eq. (19) can be expressed in terms direct strains:
S = (H1 21H2 ) d11 = k1 d11
(22)
Likewise, in the transverse direction,
S = (12 H1 H2 ) d22 = k2 d22
(23)
Solving Eqs. (22,23) yields:
H1 =
21k2 + k1
k2 + 12k1
H2 =
1 1221
1 1221
(24)
where k1 and k2 are the slope of experimental calibration curves for TSA signal
vs 11 and TSA signal vs 22 . For a general plane stress case, the TSA signal can be
expressed as the general case in Eq. (19).
2.3
IR Test Setup
A DeltaTherm DT1500 thermoelasticity measurement system was used to acquire the

thermal measurements. The DeltaTherms infrared array detector synchronized with
the applied cyclical loading enables the detection of the transient thermoelastic effect.
The infrared detector acts as a transducer, which converts the incident radiant photon
flux into electrical signals. A lock-in analyzer (a signal-processing unit) extracts the
thermoelastic information from the detectors output signal by using the reference
signal from the loading device. The TSA-IR system uses the reference signal to reject
any non-stress related thermal emissions. The frequency should be high enough to
prevent heat transfer due to stress gradients during the load cycle. The Delta-Therm
has a thermal resolution of at least 1 mK for image exposure times of one minute
or less. Figure 2 shows a schematic for the testing setup testing setup, and Figure 3
33
shows the experimental setup for the thermomechanical calibration. The IR camera
captures images at rates of more than 400 frames per second. The applied load
signal is used to integrate synchronized TSA images that correspond to peak values
of loading. The integration of the captured images is a temporal smoothing process
performed over a specified period. In this study, a period of 1-2 min was used. The
cyclic load was applied using an MTS 810 servo-hy draulic test system with a 22.2
kN(50 kip) capacity. The accuracy of the recorded strains is within 50 microstrostrains
and the load is within 0.22 kN (50lb).
Figure 2: Schematic of TSA Setup for thermomechanical calibration
34
Figure 3: Setup for Thermo-mechanical calibration
35
2.4
Validation of TSA Technique on Prepreg and Pultruded

Composites
The proposed TSA method was calibrated for Eglass/polyester pultruded, S2glass/epoxy
prepreg, and Carbon/Epoxy FRP composite systems. The FRP coupons were tested
with the load applied parallel and transverse to the major stiffness axis. The mechanical properties of the tested coupons are shown in Table 1 The Eglass tension coupons
were 25.4 mm (1.00 in.) wide, had 152.4 mm (6 in.) ungripped length and a thickness
of 12.2 mm (0.48 in.). The S2glass tension coupons were 25.4 mm (1.00 in.) wide,
had 152.4 mm (6 in.) ungripped length. A uniform thickness of 2.26 mm (0.089 in)
was achieved for the [05 /90/05 ] system using an autoclave, and a uniform thickness
of 1.78 mm (0.070 in) was achieved for the quasi-isotropic system using an autoclave.
The unidirectional Carbon/epoxy tension coupons were 25.4 mm (1.00 in.) wide, had
152.4 mm (6 in.) ungripped length and a thickness of 1.16 mm (0.046 in.). Biaxial
strain gages were oriented in the material directions to measure the surface strains.
The TSA effect was calibrated for the linear elastic region of the stress-strain response
curve (< 0.25) . Time effects on the stress-strain response during each cyclic load
application were not considered significant and the material was assumed to be under steady-state conditions. The effective properties of the Pultruded E-Glass/Epoxy
system in Table 1 were obtained by Haj-Ali and El-Hajjar [48]. For more details on
the stress-strain response of these systems, see Johnson [73].
Table 1: Effective Elastic Properties
System
E11, GPa (Msi) E22, GPa (Msi)
Pultruded E-Glass/Epoxy
17.1 (2.484)
9.96 (1.444)
Carbon/Epoxy
120. (17.46)
10.5 (1.235)
S2Glass/Epoxy[05 /90/05 ]
49.4 (7.168)
17.3 (2.830)
S2Glass/Epoxy[45/0/45/90]s 26.7 (3.879)
26.7 (3.879)
36
G12 , GPa (Msi)

3.5 (0.507)
6.9 (1.010)
13.6 (1.976)
17.5 (2.536)
v12
0.283
0.333
0.229
0.307
2.5
Thermomechanical Calibrations for Various FRP Composites
A calibration shown in Figure 4 was repeated in this study for the pultruded system
used by Haj-Ali and El-Hajjar in [48]. The thermomechanical constant obtained in
this study is 9.8 % lower than the value obtained by their group, k = 5.5973x104
camera units/(in/in). The difference may be due to differences in the numerical
algorithms used to obtain TSA averages for the specimen, ways of processing the
strain data to calculate, degrad ation of the camera sensors over time, or variablity
and differences in manufacturing of the pultruded Eglass/epoxy plates.
Figure 4: Thermo-mechanical calibration for E-glass/epoxy

Next, a unidirectional carbon/epoxy system is investigated. Notice in the calibration in Figure 5 that the 90 specimens show a slightly different slope than the
0 specimens. One should be aware of verifying computational models 2 distinct calibration constants. The thickness of the surface coat should be considered in this
37
case. It may be necessary to characterize certain layups with 2 calibration constants

especially if a proper isotropic coat is not applied. Also notice in Figure 5 that a small
number of data points were taken at high frequencies for the larger change in strain
values. This is because the MTS machine was not tuned for each frequency. The
error in the change in force applied to the specimen was neglected since the relation
is in terms of the change in the sum of measured strains and TSA signal.
Figure 5: Thermo-mechanical calibration for Uncoated Carbon/Epoxy
Figure 6: Schematic showing the lay-up for [05/90/05 ]

38
Notice in the calibration in Figure 7 that the 90 specimens show a different slope
than the 0 specimens. The thickness of the surface coat should be considered in this
case. It is necessary to characterize this layups with 2 calibration constants or apply
an isotropic coat. An additonal epoxy coating of approximately 0.3 mm (0.12 in) was
cured at room temperature to the surface of the [05 /90/05 ] layup. This surface coat
was then sprayed with flat black paint and the calibration from Figure 4 was repeated,
and the result is shown in Figure 8. As you can see the isotropic surface coat allows
the system to be characterized by a single constant that relates the Thermoelastic
signal to the sum of the normal strains.
Figure 7: Thermo-mechanical calibration for uncoated S2glass/Epoxy for the

[05 /90/05 ] Lay-up
The longitudinal stiffness of the [05 /90/05 ] system was measured to provide validitiy to the assumption of the negligible stiffness difference in adding the epoxy
coating. Table 1 shows the values of the standard [05 /90/05 ], while the properties
obtained from the epoxy coated samples are E11 = 42.3 GPa (6858 ksi) and E22 =
20.4 GPa (2957 ksi), respectively. These values are within five percent different than
that obtained from uncoated specimens. This change of stiffness due to the epoxy
39
coat can be considered within the experimental error range and justifies using the
proposed coat for TSA measurements. The calibration constant labeled All Fiber
Directions, K = 3.2477x104 , will be used to validate different coated material systems to show that the epoxy coating is being calibrated in Figure 8. Figure 9 shows
the themomechanical calibration repeated for a quasi-isotropic S2-glass/Epoxy with
a [45/0/45/90]s layup scheme. This layup was chosen because the surface ply is
orientated at -45 degrees from the loading axis. This surface ply tends to change the
orientation of the thermoelastic signal with the angle of the surface ply. Two different
quasi-isotropic lay-ups should theoretically have globally identical strain field. Bakis
et al. [14] showed that two quasi-isotropic laminates resulted in different measured
thermoelastic emission patterns for different stacking sequences. Adding an isotropic
surface coat to the quasi-isotropic laminate acts as a strain witness. Two different
quasi-isotropic lay-ups with the same added isotropic surface coating should theoretically have the same thermoelastic emission pattern for different stacking sequences.
The thickness of the surface coat is the subject for future research.
Figure 8: Thermo-mechanical calibration for coated S2-glass/epoxy

40
Figure 9: Thermo-mechanical calibration for quasi isotropic S2-glass/Epoxy
41
2.6
Validation under a Multi-axial state of stress
A standard open-hole specimen ASTM D 5766/ D 5766M was chosen for validation of
the calibration under a multi-axial state of stress. The specimen dimensions were 1.5
in wide, with a centered 0.25 in hole, and an ungripped length of 6 in, and a thickness
of 0.046 in. The specimen was cut so that the major stiffness axis is the axial direction
of the loaded geometry. The specimen was sprayed with flat black paint and the force
on the specimen was recorded as the system was cyclically loaded. This change in
force was later used in computational models. A quarter symmetry finite element
model of the open-hole unidirectional carbon/epoxy system with material properties
provided in Table 1 was chosen for computational analysis. The specimen geometry
and finite element mesh are shown in Figure 10. The model consisted of 30,000
of Abaquss CPS8R elements with symmetric boundary conditions on the quarter
geometry. A uniformly distributed load (half of that applied to the actual specimen)
was applied to the model, and the sum of the in-plane strains along the horizontal
and vertical axes through the center of the hole were plotted in Figure 12 through 15
and 17 through 24. The results were used to how the experimental results compared
to finite element and to show the effect of adding a surface witness layer.
42
Figure 10: Geometry of notched specimen used to verify the TSA technique and a
very refined quarter-model FE mesh used to model the notched specimen
43
A TSA image of an un-coated unidirectional carbon/epoxy open-hole specimen

was taken, and the result is shown in Figure 11. Notice in this case a small amount
of damage resulted upon drilling the unidirectional composite, and some loose fibers
are seen at the hole edge on the left side of the image of the un-coated specimen in
Figure 11. Future open-hole specimens were drilled with a brad point drill bit at a
low drill speed (600 rpm) as suggested by Mackin et al. [93].
Figure 11: TSA images of open hole specimens of Carbon/epoxy

Figures 12 and 15 show vertical and horizontal line interrogations drawn on the
TSA data for a 0 degree specimen through the center of the hole using the 0 degree
calibration constant (k = 2.1415x104 as shown in Figure 5) to scale the data. The
vertical line interrogation shows thermoelastic data for a 38.1 mm (1.5 in) line through
the center of the hole as shown in Figure 10. A reasonable accuracy was achieved
for model verification using the 0 degree calibration constant with a 0 degree open
hole coupon. To verify the proposed calibration, the surface of the specimen must
be coated with epoxy or alternative isotropic coat, and the experimental calibration
must be re-examined so that a single calibration constant is achieved. In this case
however, the 0 degree calibration constant was used because a small error in
would result from an experimental test where the TSA values are less than 2000
camera units. So within this range of the calibration, a single constant was used to
44
verify a computational model.
Figure 12: Horizontal line interrogation of open-hole geometry for verification of

thermomechanical calibration of un-coated unidirectional carbon/epoxy
45

thermomechanical calibration of coated unidirectional carbon/epoxy
Figure 14: Vertical line interrogation of open-hole geometry for verification of thermomechanical calibration of coated unidirectional carbon/epoxy
46
Figure 15: Vertical line interrogation of open-hole geometry for verification of thermomechanical calibration of un-coated unidirectional carbon/epoxy
For validation of the S2-glass/epoxy system the same geometry shown in Figure
10 was chosen for validation of the calibration under a multi-axial state of stress. The
major stiffness axis is the axial direction of the loaded geometry. The lay-up of the
two S2-glass/epoxy systems are quasi-isotropic and a [05 /90/05 ] lay-up as shown in
Figure 6. In the case of the [05/90/05 ] lay-up, one 90 ply is placed in the middle of the
lay-up to provide additional stability to the system. A uniform thickness of 0.089 in
was achieved for this system using an autoclave. A uniform thickness of 0.0662 in was
achieved for the quasi-isotropic system using an autoclave. The specimen was sprayed
with flat black paint and the force on the specimen was recorded as the system was
cyclically loaded. This change in force was later used in computational models. The
same quarter geometry finite element mesh was used for the computational analysis
of this system with material properties provided in Table 1.
47
The computational results for two loading cases with the corresponding experimental loading cases are plotted in Figures 17 and 21 for the [05 /90/05 ] lay-up.
These experiments take into account the thermomechanical constant provided in Figure fig:cals2, k = 3.8195x104 . The full-field infrared signal can be used to verify
various specimen geometries and material systems with some accuracy. Notice the
uniformity of the TSA signal in Figure 16. Overall, the results illustrate the accuracy
of the proposed method in determining the full-field strain invariant in various FRP
systems. In both cases only a coating of flat black paint was applied to the specimen
surface. Improved results may be obtrained by adding a layer of continuous fiber mat
material or epoxy to the surface. The 45 layer of S2-glass/epoxy on the surface of the
Quasi Isotropic layup shows that the surface ply plays a large role in the thermoelastic
response.
Figure 16: TSA images of open hole specimens of S2-glass/epoxy
48

thermomechanical calibration of un-coated S2-glass/epoxy for the [05 /90/05 ] lay-up

thermomechanical calibration with epoxy coating of S2-glass/epoxy for the [05 /90/05 ]
lay-up
49

thermomechanical calibration of un-coated quasi-isotropic S2-glass/epoxy lay-up
50
The lack of fit for the prediction shown in Figure 19 can be explained due the
multi-axial state of stress in the material system of the top layer, i.e. normal and
transverse stresses. Previous predictions from this type of calibration were done for
the 0 top layer with predominately axial stresses. This may also explain the good
quality in those cases.

thermomechanical calibration with epoxy coating of quasi-isotropic S2-glass/epoxy
lay-up
51
Figure 21: Vertical line interrogation of open-hole geometry for verification of thermomechanical calibration of un-coated S2-glass/epoxy for the [05 /90/05 ] lay-up
Figure 22: Vertical line interrogation of open-hole geometry for verification of thermomechanical calibration with epoxy coating of the S2-glass/epoxy for the [05 /90/05 ]
lay-up
52
Figure 23: Vertical line interrogation of open-hole geometry for verification of thermomechanical calibration for un-coated S2-glass/epoxy with a quasi-isotropic lay-up
Figure 24: Vertical line interrogation of open-hole geometry for verification of thermomechanical calibration with epoxy coating for S2-glass/epoxy with a quasi-isotropic
lay-up
53
CHAPTER III
MECHANICAL BEHAVIOR OF JUTE HYBRID

BIO-COMPOSITES
3.1
Introduction
This chapter utilizes TSA as before, but this section discusses analytical work and
towards investigating nonlinear behavior in Jute Hybrid Bio-Composites where TSA
techniques are utilized for validation of the computational modeling approaches. An
experimental and analytical study is carried out to investigate the mechanical behavior of biocomposites in the form of pultruded layers manufactured with Jute bio-fibers,
combined with unidirectional roving E-glass, and embedded in a polymeric matrix.
Stress strain curves are generated for these dually reinforced systems in transverse, axial and shear modes to calibrate the nonlinear parameters for computational models.
Two macro and and one micromechanical constitutive models are proposed to characterize the nonlinear orthotropic behavior of these material systems and implement
them within a finite element (FE) code. The models are validated under a multi-axial
state of stress by full-field stress/strain analysis via digital image correlation (DIC)
and Thermoelastic Stress Analysis (TSA) of open-hole specimens. While these hybrid composites have relatively low modulus, their material consistency lend them to
be used in many structural applications. Hybrid bio-composites are environmentally
friendly and can provide a sustainable alternative to existing engineering materials in
several applications. Section 2 of this manuscript introduces the bio-composites used
in this study along with their composition and overall linear mechanical properties.
Section 3 of this study formulates three nonlinear constitutive material models that
can be used to predict the multi-axial stress-strain behavior of bio-composites. Two
54
of these nonlinear anisotropic homogeneous models assume that the material is under a state of plane stress. The third is a Nine-Cell nonlinear 3D micromechanical
model using a special case of the more general High Fidelity Generalized Method of
Cells (HFGMC) proposed by Aboudi [101] and modified by Haj-Ali and Aboudi [58].
The formulation for the HFGMC model is also given in this section. The models
from Section 3 are calibrated and validated in the third part of this study with Thermoelastic Stress Analysis (TSA) and Digitial Image Correlation (DIC) for a full-field
stress-strain analysis of plates with a hole. A more detailed description of techniques
used for validation of computational analysis using TSA is available in the literature
on static, fatigue, and fracture behavior of FRP composite materials for quantitative
comparisons in the Introduction.
3.2
Jute Hybrid Bio-composite Systems
The hybrid composite system for this study is composed of continuous Jute yarns
running the length of a flat pultruded plate sandwiched between glass chopstrand mat
surface layers. The glass mat reinforcement provides some transverse reinforcement.
The averaged properties of the material system used in this study are shown in Table
2.
Table 2: Description of Jute hybrid bio-composite material system
Material Constituent
Volume Fraction
Jute Yarn
0.360
E-Glass Chopstrand Mat 0.1325
Matrix
0.5075
To get an estimate of the fiber volume fractions, photo microscopy was used to
measure the cross-sectional areas of the Jute fibers and chopstrand glass mat layers.
Notice in Figure 25 that the Jute fibers are a woven bundlea non-circular shape.
The Jute fibers are visibly darker than the surrounding matrix and the chopstrand
mat glass layers, so this fact is used to apply image analysis algorithms to segment
55
the fibers from the matrix. In Figure 26, the first steps are taken to create a binary
image where segmented areas can be measured.
Glass Mat Layer
Jute Fiber
Jute Fiber
Matrix
Glass Mat Layer
0.1065 in
Figure 25: Hybrid Jute/Glass composite with chopstrand mat surface layer
6
0.1065 in
Figure 26: Color similarity algorithms used to identify and segment Jute yarns from
background
56
3.3
Nonlinear Constitutive Models for Bio-Composites
Many experimental and analytical studies have been focused on the nonlinear response of laminated composite materials. Macro-level theories are formulated to
characterize the nonlinear response by idealizing the composite as an anisotropic homogeneous medium. Petit [102] proposed an incremental approach for the analysis
of symmetric laminates under uniform membrane loading. In their approach, the
elastic constants of a lamina are determined as a function of the lamina strains. A
lamina failure is identified when any of its strain components exceeds the experimentally obtained limiting strain value. Once a failure mode is detected in the lamina,
the corresponding tangent stiffness is set to a high negative value in order to achieve
stress unloading in this direction. Their analyses predicted the overall shape of the
nonlinear response quite well, while the laminate ultimate failure loads were predicted
with acceptable accuracy. Hahn and Tsai [55] used the complementary energy density polynomial function, for a lamina under a plane-stress state, to derive a nonlinear
stress-strain relation for laminated composites. An additional fourth-order term of
the axial-shear stress is added to the polynomial function. This term represents the
nonlinear shear strain. Interaction terms between the three stress components were
neglected. Hahn [54] extended this work to model the nonlinear behavior of laminates, and considered the effect of material nonlinearity on the buckling load of a
symmetric laminate. Hashin et al. [67] formulated a new nonlinear constitutive model
where inelastic transverse and axial shear strains exist in a lamina under plane stress
conditions. Each inelastic strain component is an independent quadratic function in
terms of the stress invariants raised to a general power. Ramberg-Osgood representation of nonlinear stressstrain curves was used to express the nonlinear axial-shear
and transverse stressstrain relations. Predicted nonlinear behavior compared well
with experimental results for different laminated composite materials. Jones and
57
Nelson [77] proposed a material model based on strain energy that accounts for nonlinear behavior under biaxial states of stress. This model can only be applied to
multi-axial stress states where the strain energy is lower than the maximum strain
energy generated in the uni-axial test results. Since this is often the case when the
composite is subject to general multi-axial stress states, different approaches were proposed to overcome this difficulty, Jones and Morgan [76], Abu-Farsakh [7]. Amijima
and Adachi [10] introduced a simple technique to represent the shear nonlinearity
using piecewise linear segments. Sandhu [112] introduced a technique for solution
of the nonlinear laminate equations using piecewise cubic spline functions to represent the lamina uni-axial test data. He also proposed a predictor-corrector iterative
method. Nahas [99] presented a similar predictor-corrector technique in which the
basic uni-axial stress-strain relations are allowed to have a general nonlinear representation. Kuppuswamy et al. [86] used the Richard and Abbott [106] representation
for the nonlinear uni-axial stress-strain relations in a three-dimensional finite element
analysis. However, their study uses uncoupled one-dimensional stress-strain relations
which do not reflect interaction effects. Pindera and Herakovich [103] and Mathison
et al. [96] derived a nonlinear plane stress constitutive model using orthotropic endochronic theory. This theory is based on irreversible thermodynamics with internal
variables. Dvorak and Rao [45] proposed a plasticity theory for fibrous composites
under axisymmetric deformation. Plastic dilatation and deformation of the composite in the fiber direction were accounted for. A yield function was formed using the
stress invariants of a transversely isotropic material and simple hardening and flow
rules were derived. Griffin [53] extended Hills anisotropic plasticity theory, and used
the associated plasticity flow rule to determine the plastic increment of strain. A
Ramberg-Osgood uni-axial stress-strain relation was used to model nonlinear hardening. Sun and Chen [119] developed a one parameter orthotropic plasticity model for
thermoplastic and metalmatrix composite laminates in plane stress. In this model,
58
one-parameter plastic-potential scalar function is proposed. Only transverse and axial shear stresses are involved in the plastic potential; therefore, plastic deformation
in the axial direction of the lamina is neglected. In addition, a power law relation
between effective plastic strain and the effective stress is proposed; as a result, a total
of three material parameters are needed to fully describe the lamina plastic behavior. Sections 3.1 and 3.2 of this manuscript describe the formulations for nonlinear
orthotropic macromechanical models used for bio-composites. Section 3.3 of this
manuscript describes the formulation of a multiphase micromechanical constitutive
model used for capturing nonlinear anisotropic behavior in bio-composites. Section 4.
presents experimental calibration and validation of the models implemented within
ABAQUS [2] finite element program as formulated in Section 3.
3.3.1
Anisotropic Deformation Theory (ADT) Based Model
The ADT model presented here is similar to Hashins [67] anisotropic deformation
theory with the addition of inelastic behavior for the 11 stress in the fiber direction.
Decompose the strain into elastic and inelastic parts:
ij = eij + Iij
(25)
Similar to the isotropic case: Iij = f(J2 )Sij We assume:

Iij = f(L)ij
2
2
2
L = 211 11
+ 22222
+ 266 66
(26)
(27)
(28)
One dimensional Ramberg Osgood relations in the 22 direction:

22
I22

n 1
1
|2| 2
=
22 + 2
2
E22
02
"

n2 1
2 # n221
|2 |
|2 |
= 2
22 = 2
22
02
02
59
(29)
(30)
taking the general form: I22 = f22 (L)22 then,

2
f22 (L) = f22(22222
) = 2
"
|2 |
02
2 # n221
(31)
and can be defined as:

2
222 = 02
; f22 = 2 L
n2 1
2
2
211 = 01
; f11 = 1 L
n1 1
2
2
266 = 06
; f66 = 6 L
n6 1
2
(32)
(33)
(34)
Therefore:
n1 1
1
12
11
22 + 1L 2 11
E11
E11
n1 1
12
1
11 +
22 + 2L 2 22
=
E11
E22
n6 1
1
=
12 + 6L 2 12
G12
11 =
(35)
22
(36)
12
(37)
If the Ramberg Osgood form is:

1
o
=
+
E
E
n
I22 =
202 22 n2
2
(
) =
E22 02
E22
I22 =
1
f22(L)22
E22
2
f22(L) = f22(22222
) = 2
(38)
"
|22|
02
2 # n221
22
(39)
(40)
"
|22|
02
2# n221
(41)
Therefore:
1
12
1 n1 1
11
22 +
L 2 11
E11
E11
E11
12
1
2 n1 1
=
11 +
22 +
L 2 22
E11
E22
E22
1
6 n6 1
=
12 +
L 2 12
G12
G12
11 =
(42)
22
(43)
12
60
(44)
Recall:
L =
11
=
1
11
=
2
11
=
12
1
E11
2
11
01
n1 1
1
L 2
E11
12
E11
22
02
2
12
06
2
n1 3
1 n1 1
L
( 2 )L 2
1
E11
11
1 n1 1 n123 L
L
E11 2
22 1
1 n1 1 n1 3 L
L 2 12 1
E11 2
L
211 L
222 L
212
= 2 ;
= 2 ;
= 2
11
01 22
02 12
06
(45)
(46)
(47)
(48)
(49)
The tangent stiffness is therefore:
d22
d11
d22
d22
d22
d12
d12
d1
d12
d2
d12
d6
3.3.2
12
E11
=
=
1
E22
2 n2 1 n223 L
L
E22 2
1 2
n2 1
2
L 2
E22
n2 3
2 n2 1
L
( 2 )L 2
2
E22
2
(50)
(51)
n2 3
2 n2 1
L
( 2 )L 2
2
E22
12
(52)
n6 3
6 n6 1
L
( 2 )L 2
6
G12
1
(53)
n6 3
6 n6 1
L
( 2 )L 2
6
G12
22
(54)
1
G12
n6 1
6
2
L
G12
n6 3
6 n6 1
L
2
(
)L
G12
2
12 12
(55)
Anisotropic Potential Theory Model (APT)
The APT model is an energy based nonlinear orthotropic model similar to HahnTsais [55] nonlinear orthotropic plane stress model with the addition of inelastic
terms in the transverse and axial directions.
61
The energy function can be defined as:

W = W(11, 22, 12) = W(1, 2, 6 )
(56)
The linear stress vs. strain relations take the form:

12
1
11
22
E11
E11
12
1
=
11
22
E22
E22
1
=
12
G12
11 =
(57)
22
(58)
12
(59)
The complementary energy function,W , is the sum of the linear and nonlinear cases.
W is the strain energy density.
W = Wl + Wnl
(60)
1
W = ij ij
2
(61)
Expanding on the strain energy density from index format:

2Wl = 1111 + 2222 + 212 12 = 1111 + 22 22 + 1212
2Wl =
1
12
12
112
1122
1122 +
E11
E11
E11
1
212
112
1122 +
2Wl =
E11
E11
1
222 +
E22
1
222 +
E22
1
122
G12
1
122
G12
2Wnl = F1(11) + F2 (22) + F6(6)
(62)
(63)
(64)
(65)
To account for nonlinearity the 1D Ramberg-Osgood form is implemented in this

model.
1
||
o
1

1
||
= +
E
o
1
= +
E
62
(66)
(67)
The total strain is then the sum of the linear and non-linear parts:
11
22
12
1
|11| 1
11
o1

1
1
|22| 2
12
11 +
22 + 2
22
=
E11
E22
o2

1
1
|12| 6
=
12 + 6
12
G12
o6
1
12
=
11
22 + 1
E11
E11
(68)
(69)
(70)
Iterative Implementations with Residual Given: (11, 22, 12) find (1 , 2, 6)

R1
R2
R3
1
|11| 1
11 11 = 0
o1

1
12
1
|22| 2
=
11
22 + 2
22 22 = 0
E11
E22
o2

1
1
|12| 6
=
12 + 6
12 12 = 0
G12
o6
12
1
=
11
22 + 1
E11
E11
R
CT
CT =
1
E11
+ 1 1
12
E11
|11 |
o1
1 1
1
1
E22
+ 22
2 1
1
G12
R() = R(1, 2 , 6) = 0

R
(i)
( (i+1) (i)) = 0
R()=R( ) +
(i)

1

R
(i+1)
(i)
{R( (i))}
(
)=
(i)

= CT R( (i))

(i)
0

+ 6 6
Using taylor expansion:
63
(73)
0
|22 |
o2
(72)
(74)
12
E11
(71)
|12 |
o6
6 1
(75)
(76)
(77)
(78)
(79)
In general:
11
22
12

1
12
=
+ f1 (11) 11
+ f21 (22) 22
E11
E11

1
12
+ f12(11) 11 +
+ f2 (22) 22
=
E11
E22

1
=
+ f6 (6 ) 6
G12
(80)
(81)
(82)
Note that assuming f12 = f21 = 0 implies no nonlinear poissons effects.

3.3.3
New Formulation for the Nine-Cell Micromechanical Model (NineCell)
The newly developed micromechanical model for the lamina is shown in Figure 27.
Previous implementations include a two, four, and multi-cell models developed by
Haj-Ali et al. [56, 57, 5961, 81] for nonlinear analysis of E-glass/vinylester pultruded
composites validated under different multi-axial states of stress with and without
damage under tension, compression, bending and off-axis. In the Nine-Cell model a
unidirectional lamina is represented by a 3-by-3 matrix of constituents with rectangular cross-sections representing the binder and long fibers which account for the 3D
mechanical response. The nine cells (subcells) within the Unit Cell (UC) are given
properties of the Jute fiber, glass chopstrand mat, and matrix constituents. The
micromechanical framework allows a description of the subcells and material constituents that can be used to account for damage mechanisms and overall nonlinear
and damage softening behavior. The mechanical response of the UC is determined
by satisfying the traction continuity and strain compatibility constraints between the
subcells.
The Nine-Cell model is similar to the is a simplified version of the more general
case of the High Fidelity Generalized Method of Cells given by Haj-Ali [58], and the
derivation here includes the incremental formulation with integration and correction
schemes as described in Haj-Ali [56].
64
Figure 27: (a) Idealized medium with periodic arrays of long fibers (rectangular crosssections) showing a quarter UC (bold line), (b) Nine-cell micromodel of rectangular
UC
65
The unidirectional composite shown in Figure 27 consists of long fibers embedded

in the matrix system. It is idealized as doubly periodic array of fibers with rectangular
cross section. Taking into account symmetry, a quarter UC is used here that consists
of nine subcells is used in the formulation. Instead of assigning interface subcells as
discussed in [56], all of the cells are used to capture the nonlinear behavior without
damage. The first subcell is a fiber constituent, while subcells 2, 3, and 4 represent
the matrix constituents. The other subcells, 59, represent the interfaces. The long
fibers are aligned in the x1 direction. The other cross-section directions are referred to
as the transverse directions. The x3 direction is called the out-of-plane axis or lamina
thickness direction. The nine-cell micromodel is formulated by describing the relations
between the average stress and strain components in the different subcells. The new
derivations are carried out assuming the unit cell is subject to uniform average strain
rate for each time step. Micromechanical modes of deformations, mode 1 to mode 6,
are used to describe the relations between the average stress and strain components
of the different subcells.
The notations for the stress and strain vectors, defined in this section, are:
()
= [11, 22, 33, 12 , 13, 23 ]T i = 1, ..., 6
(83)
()
= [11, 22, 33, 12 , 13, 23 ]T = 1, ..., 9
(84)
i
i
where, () denotes the subcell number. The total volume of the UC is taken to be
equal to one to simplify the fiber and matrix volume calculations for estimates of the
overall mechanical response. The relative volume of each of the nine subcells, denoted
by v, can be easily calculated from Figure 27. The axial strains in the subcells are
the same and equal to the overall strain in the UC. This is basic first order axial
displacement continuity and similar to the well-known Voigt model. Therefore, the
longitudinal relations (mode 11) are:
66
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
d1 = d1 = d1 = d1 = d1 = d1 = d1 = d1 = d1 = d1
(1)
(2)
(3)
(9)
v1d1 + v2 d1 + v3d1 + ... + v9d1 = d

1
(85)
(86)
Consideration of the interfaces normal to the x2 direction, yields the traction

continuity conditions for modes 22 and 12, respectively. The corresponding strain
compatibility conditions for these modes follow from considering the contribution of
subcells (1), (2), and (5), the assembly of subcells (3), (4) and (6), and the third row
of subcells (7), (8), and (9). These relations are expressed for mode 22 as:
(1)
(2)
(7)
(87)
(2)
(4)
(8)
(88)
(5)
(6)
(9)
(89)
v1d2 + v2 d2 + v5d2 = (v1 + v2 + v5)d2
d2 = d2 = d2
d2 = d2 = d2
d2 = d2 = d2
(1)
(2)
(5)
(90)
(3)
(4)
(6)
(91)
v3d2 + v4 d2 + v6d2 = (v3 + v4 + v6)d2

(7)
(8)
(9)
v7d2 + v8d2 + v9 d2 = (v7 + v8 + v9)d2
(92)
For mode 12 or mode 4, the relations are:
(1)
(2)
(7)
(93)
(2)
(4)
(8)
(94)
(5)
(6)
(9)
(95)
v1d4 + v2 d4 + v7d4 = (v1 + v3 + v7)d4
d4 = d4 = d4
d4 = d4 = d4
d4 = d4 = d4
(1)
(3)
(7)
(96)
(2)
(3)
(8)
(97)
v2d4 + v4 d4 + v8d4 = (v2 + v4 + v8)d4

(5)
(3)
(9)
v5d4 + v6d4 + v9 d4 = (v5 + v6 + v9)d4
67
(98)
Consideration of the interfaces normal to the x3 direction yields the traction continuity conditions for modes 22 and 12, respectively. The corresponding strain compatibility conditions for these modes follow from considering subcells (1) and (3),
and subcells (2) and (4), respectively, to be connected in series. These relations are
expressed for mode 33 as:
(1)
(2)
(7)
(99)
(2)
(4)
(8)
(100)
(5)
(6)
(9)
(101)
v1d3 + v2 d3 + v7d3 = (v1 + v3 + v7)d3
d3 = d3 = d3
d3 = d3 = d3
d3 = d3 = d3
(1)
(3)
(7)
(102)
(2)
(3)
(8)
(103)
v2d3 + v4 d3 + v8d3 = (v2 + v4 + v8)d3

(5)
(3)
(9)
v5d3 + v6d3 + v9 d3 = (v5 + v6 + v9)d3
(104)
For mode 13, the relations are:
(1)
(3)
(7)
(105)
(2)
(4)
(8)
(106)
(5)
(6)
(9)
(107)
v1d5 + v3 d5 + v7d5 = (v1 + v3 + v7)d5
d5 = d5 = d5
d5 = d5 = d5
d5 = d5 = d5
(1)
(3)
(7)
(108)
(2)
(4)
(8)
(109)
v2d5 + v4 d5 + v8d5 = (v2 + v4 + v8)d5

(5)
(6)
(9)
v5d5 + v6d5 + v9 d5 = (v5 + v6 + v9)d5
(110)
Finally, in the transverse shear mode, the traction continuity at all the interfaces
between the subcells must be satisfied. Since the subcells have a constant stress, the
relations for the transverse shear mode, mode 23 or mode 6, are expressed as:
68
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(111)
v1d6 + v2d6 + v3d6 + ... + v9 d6 = d6
(112)
d6 = d6 = d6 = d6 = d6 = d6 = d6 = d6 = d6
(1)
(2)
(3)
(9)
It is instructive to recall the definition for the UC average strain:
9
1 X
()
V di = di i = 1, 2, ..., 6
V =1
(113)
Note that all the strain compatibility relations in Equations (2)-(7), explicitly
satisfy Equation 113. The only independent relation that is not identically satisfied in
Equation 113 is the transverse shear case, i = 6. Therefore, this relation is added and
considered as a compatibility relation. Next, the incremental stress-strain relations
are expressed for each subcell as:
()
di
()
()
= Cij dj
i = 1, 2, ..., 9
(114)
The UC average stress increment is:
9
1 X
()
V di = di i = 1, 2, ..., 6
V =1
(115)
For a given average strain increment imposed on the UC, Equations 85 to 115
provide a system of 114 equations. The 114 unknowns are:
()
()
di , di , di
= 1, 2, ..., 9; i = 1, 2, ..., 6
(116)
These unknowns can be further reduced by using the strain compatibility relations
and divide the variables into two independent and dependent sets. The combined set
69
of equations that describe the strain compatibility and the traction continuity equations (micromechanical constraints) can ultimately be written in a general incremental
form as:
dR = i (d , d(), d, V(), = 1, 2, ..., N) = 0; i = 1, ..., a
(117)
Equation 117 is the linearized incremental form of the micromechanical relations.

It can be solved to generate the tangential incremental update and evaluate the
current strain-interaction matrices for the subcells assuming linear material behavior
in the subcells along with no damage within the current step. The incremental form
of the traction-continuity relations along with its total form, i.e., dR and its total
form R , can be used to iteratively solve for the micromechanical equations since
the incremental update of the variable from the linearized relation will almost always
generate nonlinear response at the constitutive levels of the matrix and interface
subcells. The incremental form of the stress-strain relations in all the subcells can be
generally expressed in incremental fashion as:
dR = j (C , d(), d, V(), = 1, 2, ..., N) = 0; j = 1, ..., b
(118)
At any given state of deformations, the linearized micromechanical equations can

be arranged in a general compact incremental form as
6N 6N
d(1)
(2)
d
...
...
...
d(N )
6N 1
70
= D
d

6N 6
61
(119)
The left strain vector represents the unknown strains in all the subcells. The
matrix, A, is composed form known values of the current tangent stiffness matrices
of the subcells. The right-hand strain vector, d , is the overall UC average strain.
Equation 119 can be rearranged by dividing the subcells strain components into two
dependent groups with (a) and (b) number of components, respectively to yield a
new compact form that can be solved numerically in an efficient manner. The general
structure of the linearized micromechanical equations for our micromodel is:
ab
I
A
dR
(aa)
(ab)
(a1)
ba A
bb
dR
(b1)
(ba)
(bb)
da
(a1)
db
(b1)

Da
(a6)
d

=
(61)
(120)
(b6)
The bar notation over the components of the (A) matrix denotes the new arrangement of the terms of the original matrix. This is a general formulation of any UC that
may have a different number of subcells as the special case presented here, namely
N = 9. Once Equation 119 or 120 are solved, the incremental stress in each of the
subcells and the average stress of the UC can be back calculated using the incremental
stress-strain relations. The incremental strain-concentration matrices are generally
expressed by:
B (1)
B (2)
.
.
.
B (N )
(121)
The stress analysis of a micromechanical RUC becomes a straight forward procedure as a result of this formulation. Given an average strain increment and the
history of deformations in the subcells, the strain interaction matrices are formed
using Equation 121. The strain increments are subsequently formed in each of the
71
subcells followed by the corresponding stress increments. This procedure is a linearized incremental stress analysis and will be referred to as the trial state. If only
this linearized trial analysis is used, two types of error will result at each trial increment, and will accumulate during the analysis. It is important to mention, however,
that the strain compatibility and traction continuity constraints are exactly satisfied
by the trial state, which is composed of tangential approximations. The first error
occurs in the strain increments because the strain-interaction matrices are derived
using the tangent stiffness matrices of the subcells at the beginning of the increment.
The second error occurs as a result of using the tangent stiffness to compute the
stress increment. Therefore, the proposed correction scheme is important in order
to accurately account for the nonlinear constitutive (with or without damage) material behavior (prediction) and its associated error in the incremental micromechanical
equations.
3.3.4
The HFGMC micromechanical modeling framework
The High Fidelity Generalized Method of Cells (HFGMC) is a computational nonlinear micromechanical model for multi-phase composites. A more thorough review review of the HFGMC model can be found in Aboudi [3] and in Haj-Ali and Aboudi [58]
where the modified form now includes nonlinear algorithms for implicit integration
with tailored stress correction schemes allowing for iterative error reduction through
minimization of the residual vector, and multi-scale features extending the capabilities of the model for greater efficiency in capturing the local and global load
re-distribution that is associated with nonlinear and damage effects in large-scale
structural computational finite element models consisting of doubly periodic multiphase composite media. Classical nonlinear micromechanical models of multi-phase
materials are capable of capturing the nonlinear effective response of the material but
neglect the local spatial variation of the deformation fields within the representative
72
microstructure. The HFGMC has the capability of predicting local nonlinear mechanical behavior of periodic multi-phase materials subject to remote loading through a
modeling framework that allows for a detailed description of the constituents and spatial deformations between the phases. Multiphase composites are analyzed within the
HFGMC framework with nonlinear elastic, elastoplastic and viscoplastic constituents
Aboudi2004.
The HFGMC has been used to analyze fully coupled electro-magneto-thermoelastic multi-phase composites [5], fiber-matrix debonding in metal matrix composites [19], thermomechanical analysis of internally cooled structures at elevated temperatures for aerospace engine applications [11], adhesively bonded composite joints [20],
and internal pore pressure for the analysis of foam insulation of the external tank of
the space shuttle [18]. This framework is based on the homogenization technique of
composites with periodic microstructure as depicted by the repeating unit cell (RUC)
in Figure 28. The RUC in of a composite material system, Figure 29, is divided into
an arbitrary number of rectangular cells each assigned with a material constituent
with nonlinear homogeneous properties. The size and constituent volume fraction is
defined by the indices () . The dimensions of the cell along the 2 and 3 axes are
denoted by h and l , respectively. In the present doubly periodic case of continuous
()
()
fibers, a local coordinate system (y2 , y3 ) is introduced in each cell whose origin is
located at its center, see Figure 29.
The derivation of the HFGMC shown here was developed by Haj-Ali and Aboudi
[58]. This model is used to calibrate the material system using values taken from the
literature on the constituents of the microstructure and limited experimental testing
to achieve a close estimate of the actual nonlinear anisotropic behavior, and then
the calibrated material model is used to predict the nonlinear anisotropic behavior
under a multi-axial state of stress in a particular hybrid bio-composite system for this
study. Volumetric equilibrium, traction and displacement continuities are satisfied in
73
an average sense by the HFGMC framework with a higher oder polynomial form for
the displacement field in the unit cell as shown in Figure 28.
Repeating Unit--Cell
(RUC)
Figure 28: Schematic illustration of a unidirectional periodic array in the global x2 x3

plane of multiphase composite media with its repeating unit-cell (RUC), defined with
respect to its y2 y3 local coordinate system.
74
y2
(1 N)
(N 1)
(N 1)
(1 1)
(1 N)
(1 1)
(N N)
(N 1)
(CH 1)
(CH ) y (C)
2
hC
y (H)
3
(C 1 H)
h C h C1
1
(N N) (N 1)
(1 1)
lH
l H1 l H1
(1 N)
y3
Figure 29: General arrangement of cells and their geometry and coordinate systems
for the HFGMC-RUC model. Mirrored cells-interfaces are illustrated to enforce the
periodic boundary conditions.
75
The overall displacement field, u, is a function of the global coordinate system.

Each subcell has a local displacement field, u, in a local-global coordinance system,
x and y.
= u
(x)
u
(122)
u() = u()(x, y)
(123)
In the framework of the HFGMC model for periodic composites, the displacement
vector in the cell () is given , e.g. Aboudi [6], by the higher-order polynomial form
()
()
()
()
()
u() = u + W(00) + y2 W(10) + y3 W(01)

+
1 ()2 h2
1 ()2 l2
()
()
(3y2 )W(20) + (3y3 )W(02)
2
4
2
4
(124)
where is the externally applied average strain. The coefficient variables vectors,
()
W(mn) , represent the volume averaged displacement in the case of m = n = 0, which

together with the additional higher-order terms have to be determined.
The strain vector at each cell is defined by
() {11, 22, 33, 212 , 213, 223}()
(125)
After some algebraic manipulations, it is possible to represent the strain vector in the
form
()
()
() ()
() ()
() = + P(10)W(10) + P(01)W(01) + P(20)W(20) y2 + P(02)W(02) y3
(126)
where,
P(10)
0 0 0
0 0 0
0 0 1
,
0 0 0
1 0 0
0 1 0
P(01)
76
0 0 0
0 1 0
0 0 0
1 0 0
0 0 0
0 0 1
(127)
and P(20) = 3P(10), P(02) = 3P(01) .

The average strain in the cell () can be identified as
()
()
() = + P(10)W(10) + P(01)W(01)
(128)
The strain influence matrix B() of the cell can be expressed in an incremental form
by
()
()

()
()
= I + P(10)D(10) + P(01) D(01)

(129)
with I being the identity matrix. In addition,

()
()
D
(10) = W(10) ,
()
()
D
(01) = W(01)
(130)
() are determined after the solution of the entire coupled governing equawhere D
(mn)
tions of the cells.
The resulting linear expansion of the stress vector can be expressed as
() ()
() ()
() + (10) y2 + (01) y3
() =
(131)
()
() is the average stress in the cell () and (mn)
where
are higher-order stresses
which can be directly expressed in terms of the stress moments defined by

()
S(mn)
1
=
h l
h /2
h /2
l /2
l /2
()
()
()
() (y2 )m (y3 )n dy2
()
dy3
(132)
These moments were previously employed in the original derivation of the HFGMC
(see Aboudi [6]). The current form of the stress vector is exactly equivalent to the
original derivation:
()
(10) =
12 ()
S ,
h2 (10)
()
(01) =
12 ()
S
l2 (01)
(133)
The standard pointwise equilibrium equations = 0, lead, in the framework

of HFGMC theory, to the volume-average form
()
()
L2 (10) + L3(01) = 0
77
(134)
where L2 and L3 are
L2 =
0
given by the Boolean matrices:
0 0 1 0 0
0 0 0 0 1 0
0 0 0 0 0 1
,
L
=
3
1 0 0 0 0
0 0 0 0 1
0 0 1 0 0 0
(135)
For the general case of a nonlinear material filling the cell (), the incremental form
of its constitutive relation is given by
() = C() ()
(136)
where C() is the proper instantaneous or consistent fourth-order tangent stiffness

tensor that is selected based on the considered nonlinear material behavior within the
cell, such as total deformation theory, incremental plasticity, nonlinear viscoelasticity,
and viscoplasticity. Standard forms of the fourth-order tangent or consistent tangent
tensors for these material models can be found in Simo and Hughes [116], Khan
and Huang [80] and ABAQUS [2]. For the particular Bodner-Partom viscoplastic
theory, Bodner [25], the time-dependent tangent tensor has been derived by Paley
and Aboudi [101] in the form

()
srtPrt
sij skl
()
Cijkl = ij kl + (ik jl + il jk ) 2
spq pq smn smn
(137)
where and are the Lame constants of the material filling the cell (), sij is the
deviatoric part of the stress ij in the cell, Pij is the corresponding plastic (inelastic)
strain component and ij is the Kronecker delta.
The incremental forms of the constitutive models for the different phases are
used in the linearized formulation of the HFGMC. In addition, the corresponding
total integrated constitutive forms are also employed to derive the HFGMC nonlinear
equations using total microvariables and stresses in the cells.
The incremental stress is of the form
() ()
() ()
() + (10) y2 + (01) y3
() =
78
(138)
By substituting the incremental strain given by Eq. (126) and comparing with Eq.
(138) leads to the following relations
()
()
() = C()
+ C() P(10)W(10) + C()P(01) W(01)

(10)
()
= C()P(20) W(20)
()
= C()P(02) W(02)
(01)
()
()
(139)
() is the incremental average stress in the cell. These

It should be noted that
relations are used in the equilibrium equations, Eq. (135), resulting into
()
()
L2 C() P(20)W(20) + L3C() P(02)W(02) = 0
(140)
This equation can be equivalently expressed in the compact form

()
()
()
()
A2(20)W(20) + A3(02)W(02) = 0
(141)
()
(142)
where
Ai(mn) Li C()P(mn) ,
(mn) 6= (00)
and
()
Ai(00) Li C()
(143)
Next, the traction and displacement continuity are imposed on an integral basis
over the interfaces. Consider the two interfaces between the neighboring cells ()
and (1), and () and (1) where 1 = + 1 and 1 = + 1, the displacement
continuity conditions are given by
Z l /2
u()|y() =h /2 u(1 ) |y(1 ) =h
l /2
1 /2
()
dy3
= 0
(144)
with = 1, ..., N 1 and = 1, ..., N . Similarly, the displacement continuity at the

interface whose normal is in the y3-direction is
Z h /2
u()|y() =l /2 u(1 ) |y(1 ) =l
h /2
79

()
dy2
= 0
/2
(145)
with = 1, ..., N and = 1, ..., N 1. The needed displacement periodicity conditions are
Z
l /2
l /2

u(1) |y(1) =h
(N )
1 /2
(N )
y2

()
dy3
= 0
(146)

()
dy2
= 0
(147)
=hN /2
with = 1, ..., N .
Z
h /2
h /2

u(1)|y(1) =l1 /2 u(N ) |y(N ) =l
3
N /2
with = 1, ..., N .
The two traction continuity conditions are
Z
l /2

L2 ()|y() =h
|y(1 ) =h

()
dy3
= 0
/2
(148)

L3 ()|y() =l /2 L3 (1 )|y(1 ) =l

()
dy2
= 0
/2
(149)
/2
l /2
L2
(1 )
with = 1, ..., N 1 and = 1, ..., N .

Z
h /2
h /2
with = 1, ..., N and = 1, ..., N 1. The needed traction periodicity conditions

are
Z
l /2
l /2

L2 (1) |y(1) =h1 /2 L2 (N ) |
2
(N )
y2

()
dy3
= 0
(150)

()
dy2
= 0
(151)
=hN /2
with = 1, ..., N .
Z
h /2
h /2
L3
(1)
|y(1) =l1 /2 L3
(N )
|y(N ) =l
3
N /2
with = 1, ..., N . It should be emphasized that the periodicity relations, Eq. (146)(146) and (150)-(151), are satisfied by mirroring and extending the cells near the
periodic interfaces as shown in Figure 29. Therefore Eqs. (144)-(145) and (146)-(147)
are equivalent to the periodic relations by using the proposed mirroring technique. For
example, the continuity equations can be applied to satisfy the periodicity conditions
by simply performing the shifting operation: ( = N + 1, ) = (1, ). Thus Eq.
80
(144)-(145) and Eq. (148)-(149) can be applied using the outlined mirroring procedure
to enforce the periodicity conditions as well.
The four displacement and traction continuity equations, Eq. (144)-(145) and Eq.
(148)-(149), can be expressed in the incremental form:

h2
h21
h1
h
()
()
()
(1)
(1)
(1)
W(00) W(10) + W(20) W(00) +
W(10) +
W(20)
= 0
(152)
2
4
2
4

l2
l21
l
l1
()
()
()
(1)
(1)
(1)
W(00) + W(01) + W(02) W(00) W(01) + W(02)
= 0
(153)
2
4
2
4

h ()
()
+
A2(20)W(20)
2

h1 (1 )
(1 )
(1)
(1 )
(1)
(1)
A2(10) W(10) + A2(01) W(01) +
A
W(20)
2 2(20)

(1 )
()
= A2(00) A2(00)

()
()
A2(10)W(10)
()
()
A2(01)W(01)

l ()
()
()
()
()
()
A3(10)W(10) + A3(01)W(01) + A3(02)W(02)
2

l1 (1 )
(1 )
(1)
(1 )
(1)
(1 )
A3(10) W(10) + A3(01) W(01) A3(02) W(02)
2

(1 )
()
= A3(00)
A3(00)

(154)
(155)
Let us define the vector of incremental variables of each () cell as

()

()
W(10), W(01), W(00), W(20), W(02)
(156)
Figure 30 illustrates the common interfaces between a cell () and its neighboring
cells. As previously shown, it is sufficient to consider the continuity conditions at the
two interfaces y2 = h /2 and y3 = l /2 (illustrated by bold lines in Figure 30).
81
y2
(1)
(CH )
(3)
(2) (4)
(CH 1)
(2) (4)
y3
(1)
C
C 1 C 1 (C 1 H)
H
H1 H 1
C
C 1 C 1
H
H 1 H 1
(3)
Figure 30: One-cell configuration with its neighboring cells showing its two primary
feed-forward interfaces along with the location for the stress integration points.
82
The increment of the residual vector R() for the governing equations is arranged
in the order of traction continuity, Eq. (154)-(155), the internal equilibrium equations
for the cell, Eq. (141), followed by the displacement continuity, Eq. (152)-(153),
written for the above two interfaces:

()
()
R
R , RE , Ru
()
= T1
()
D1
()
The matrix T1
(1)
X() + T2
(1 )
D2
(1 )
X(1 ) + T3
(1 )
D3
X(1)
(157)
operates on the incremental microvariable vector X() and pro-
vides the incremental traction and displacement at the two interfaces within the cell
(). Its structure is given by
A2(10) A2(01)
A3(10) A3(01)
()
T1 = 0
0
h
I
0
l
0
I
2
(1 )
Similarly, the matrices T2
h
2 A2(20)
l
A3(02)
2
A2(20)
A3(02)
h2
I
4
(1)
and T3
0
l2
4
()
(158)
operate on the incremental microvariable
vectors, X(1 ) and X(1) of the adjacent cells, respectively. They provide the
incremental traction and displacement on the other side of the two interfaces. For
(1)
example, the matrix T3
provides the incremental displacement and stress within
the cell (1 ) needed to complete the continuity conditions at the common interface
of the adjacent cell (). These two matrices are of the form
h
A2(10) A2(01) 0 2 A2(20) 0
0
0
0
0
0
( )
T2 1 =
0
0
0
0
0
2
h
h I
0
I
4 I
0
0
0
0
0
0
83
(1 )
(159)
(1 )
T3
0
0
0
A3(10) A3(01) 0
=
0
0
0
0
0
0
l1
0
I
I
2
l1
A3(02)
2
4 I
l2
(1)
(160)
The Di matrices can be viewed as the contribution to the local stresses at the considered two interfaces of the corresponding cells from the externally applied global
strain field. Their structure is given by
()
D1
A2(00)
A
3(00)
()
(1 )
D2
A2(00)
0
=
(1 )
(1 )
D3
A
3(00)
0
=
(1 )
(161)
It is interesting to note that the displacement continuity and equilibrium equations

expressed in terms of the incremental micro-variables result in a set of homogeneous
algebraic equations. This fact is useful for condensation purposes of the overall system
of equations as will be shown later.
The final system of the tangential governing equations is obtained by assembling
the contributions from all the cells by utilizing the expressions given by Eq. (157).
Figure 31 illustrates the form of the global tangential system of equations highlighting
the contribution of a general cell ().
The expanded form of the vector of variables is

(11)
(12)
(NN )
X = X , X , ..., X
(162)
Alternatively, it is possible to perform the assembly on a cell-by-cell basis where in
84
(11)
(C H)
(CH ) (CH)
(CH )
(C H)
(N CNH)
'F
'X (11)
T (CH)
2
'X (C
T (CH)
'X (CH
T (CH) T (CH
3
1
)
T (C
H)
(CH)
H)
+ D2
)
+ D3
(CH)
H) (CH
D (CH)+ D (C
D
2 +
'X (CH)
'X (CH
'X (C
)
)
H)
'X (NCN H)
Figure 31: Overall tangential system of equations for the HFGMC-RUC model indicating the contribution of cell () to the system and showing a row of complete
equations for this cell.
85
such a case all indicted matrices in this figure will be solely computed from the cell
stiffness and assembled on a column basis.
It is possible to dramatically reduce the computational effort by employing a
condensation procedure. To this end, the previously identified homogeneous and
non-homogeneous equations at the cell level can be grouped into two parts. The
re-partitioned system has the
A Au
Au Auu
form
D
=
{
}
Xu
(163)
The terms A and Auu in Eq. (163) involve the traction continuity (non-homogeneous
equations) on the one hand, and the equilibrium and displacement continuity (homogeneous equations) on the other hand. The off diagonal terms Au and Au involve
the mixed terms that emerge from re-partitioning the system of equations. The two
parts of the vector of variables, X and Xu , have similar definitions. The solution
can be represented symbolically as:

X = A1 D
D
(164)
where the square matrix A that appears on the left-hand-side of Eq. (163) has the
dimension 15N N 15N N in a non-condensed form. However, by performing the
condensation procedure the dimension of the condensed system reduces to 6N N
6N N .
The overall (global) tangential stiffness matrix defined from the effective incre = C
mental stress-strain relation,
, of the composite can be obtained by
determining the influence matrices B() defined by the incremental form of Eqs.
() and D
() can be identified from D
in Eq.
(128)-(130). The needed matrices D
(10)
(01)
86
(164) which can be partitioned as follows
()
D
()
(10)
D
D(01)
D(00)
(20)
D
(02)
D
(165)
Consequently, the effective tangential stiffness of the composite can be readily evaluated as follows (Haj-Ali [57], Aboudi [6])
N
XX
1 XX
C =
h l C() B()
h l BT ()CT ()
HL
=1
=1
=1
(166)
=1
which shows the symmetry of the effective tangential stiffness of the composite.
Next, we derive the residual vector of the governing equation for a characteristic
cell () that is needed in order to obtain the error of the nonlinear system during the
iterative solution. The total (rather than incremental) interfacial displacements and
stresses are determined from the total strains derived from the trial microvariables.
These field variables are evaluated at both sides of the neighboring cells as illustrated
in Figure 30. The locations of the integration stress points are indicated by the solid
circles. It should be noted that the displacement residuals are explicitly zero since
they do not involve stiffness variables. The residual
R
()
()
(1 )
L2
]dy3
[L2
R
()
()
(1 )
[L
]dy
3
3
2
()
()
()
R
=

L2 (10) L3 (01)
vector of cell () is given by:
()1
(1)3
L2
L2
L3 ()2 L3 (1)4
()
()
(167)
L2 (10) L3 (01)
where the first two components define the traction continuities in a total form, whereas
the third term stands for the total form of the equilibrium equations in the cell. The
87
transition from the mathematical definition of the residual vector to its numerical
form is expressed by the two parts of Eq. (167). The equilibrium equations involve
higher-order stresses that are obtained from the total stress values evaluated at the
stress integration points:
()
(10)

1
()3
()2
=
;
h
()
(01)

1
()2
()4
=
(168)
The cells residual vector R() is used in the process of assembling the overall
residual vector of the HFGMC model RHF GM C given by
RHF GM C

(11)
()
(N N )
= R , .., R , .., R
(169)
This residual vector should be equal to zero in order to solve for the nonlinear HFGMC
equations. This is numerically achieved by requiring that its norm should be sufficiently small. In the case where a mixed stress and strain combination is applied
(e.g., a uniaxial transverse loading defined by: 22 6= 0 and ij = 0 for all other
components), an additional global residual vector, RG , must be introduced, (for the
above uniaxial transverse loading, RG = {
11, 33, 23,
13,
12}), The condition that
RG = 0 should be satisfied together with RHF GM C = 0.
3.4
Calibration and Verification
Figure 32 shows the stress vs. strain relation for the Jute/chopstrand mat material
system in tension in the fiber direction. The elastic modulus is taken between 10003000 strain as discussed in the ASTM D3039 test method.
88
35
30
Stress, (ksi)
25
20
10 in
15
Experiment
Linear Fit
RambergOsgood Fit
10
E = 2885(ksi)
0.005
1 in
0.01
0.015
Strain, (in/in)
Figure 32: Jute/glass chopstrand mat system as tested in tension in the fiber direction
89
Figure 33 shows the stress vs. strain relation for the Jute/chopstrand mat material
system in compression in the transverse fiber direction. The elastic modulus is taken
between 1000-3000 strain as discussed in the ASTM D3039 test method.
(ksi)
4 in
3
Experiment 1
Experiment 2
Linear Fit
RambergOsgood Fit
2 in
E = 675(ksi)
0.005
0.01
0.015
Strain,
0.02
0.025
0.03
Figure 33: Jute/glass chopstrand mat system as tested in compression in the transverse fiber direction
Figure 35 shows the shear stress vs. shear strain relation for the Jute/chopstrand
mat material system as tested in an Iosipescu test fixture. The elastic shear modulus
is taken between 1000-3000 strain as discussed in the ASTM D5379 test method.
Table 2 shows the nominal specimen dimensions.
90
Figure 34 shows the actual specimen manufactured to the dimensions show in

Table 3 and strain gauged as discussed in ASTM D5379.
90(typ)
L/2
h
45(typ)
w
d1
d2
r (typ)
Side
L
Front
(a) Actual Specimen
(b) Schematic
Figure 34: Jute/glass chopstrand mat v-notch specimen as tested with the Iosipescu
fixture
Table 3: Nominal V-notch Specimen Dimensions

d1 = 20.0 mm [0.75 in.]
d2 = 4.0 mm [0.15 in.]
h = as required
L = 76.0 mm [3.0 in.]
r = 1.3 mm [0.05 in.]
w = 12.0 mm [0.45 in.]
91
The linear and nonlinear shear properties for constituents in the Nine-Cell model
were modified from literature by fitting the overall model behavior to the experimental
shear data as shown in Figure 35. The ADT and APT models used the mean of the
Ramberg-Osgood upper and lower bounds.
6
90(typ)
L/2
h
45(typ)
w
d1
d2
Side
r (typ)
L
Front
(ksi)
Test 1
Test 1 Ultimate
Test 2
Test 2 Ultimate
Test 3
Test 3 Ultimate
Test 4
Test 4 Ultimate
Test 5
Test 5 Ultimate
Test 6
Test 6 Ultimate
RambergOsgood Relation Upper Bound
RambergOsgood Relation Lower Bound
9 Cell
1
VNotched Test ASTM D 5379/D 5379M
=
0
0.5
o n
+
( )
G
G o
1
1.5
Strain, (%)
2.5
Figure 35: Jute/glass chopstrand mat system as tested in shear with the Iosipescu
fixture
The experimentally determined orthotropic elastic materials properties of the
Jute/Glass Chopstrand Mat system are shown in table 4.
Table 4: Unidirectional elastic material properties of Jute with chopstrand mat
E11 MPa (ksi) E22 MPa (ksi) G12 MPa (ksi) 12
Thickness mm (in)
19890 (2885)
4650 (675)
3172 (460)
0.36 2.71 (0.1065)
92
Fitting the experimental results to the Ramberg-Osgood relations =
0
+
( )n
E
E 0
results in the calibrated Ramberg-Osgood relations and paramters for Jute/glass

chopstrand mat shown in Table 5.
Table 5: Calibrated Ramberg-Osgood Relations and Parameters for Jute / Glass
Chopstrand Mat
Modulus MPa (ksi) 0 (ksi)
n
E11 =19890 (2885)
0,11 =345 (50.) 11 =1 n11 =4.0
E22 =4650 (675)
0,22 =48 (7.)
22 =1 n22 =4.0
G12 =4210 (548)
0 =30. (4.3)
12 =1 n12 =5.0
Table 6: Ultimate material properties of Jute with chopstrand mat

T 11 MPa (ksi) C22 MPa (ksi) KT L MP a m1/2(ksi in1/2)
277 (40.2)
47.0 (6.8)
22.0 (20.)
93
Table 7 gives the linear orthotropic properties used for the Jute Yarn and E-Glass
Chop strand mat material constituents. Overall orthotropy would be satisfied in the
RUC with the combination of all constituents in the RUC. Table 8 shows the nonlinear
properties used to describe the matrix using Ramberg-Osgood relations.
Table 7: Fiber linear
Fiber
E11 GPa (ksi)
Jute Yarn
46.9 (6800)
E-glass Mat 6.89 (1000)
elasetic properties used

E22 GPa (ksi) 12
3.45 (500)
0.11
6.89 (1000)
0.25
in the Nine-Cell RUC

G12 GPa (ksi) G23 GPa (ksi)
4.83 (700)
4.83 (700)
5.52 (800)
5.52 (800)
Table 8: Nine-Cell RUC Isotropic Nonlinear Matrix Material Properties (RambergOsgood Relations)
Material E11 GPa (ksi) 12 So MPa (ksi)
Matrix
4.14 (600)
0.3 1 27.6 (4)
4
Figure 36 shows a schematic representation of a Repeating Unit Cell used here
to show how the Xl and Xh parameters are used to describe the micromechanical
geometry. Notice that the subcells are rectangular and 6 parameters are needed for
the 3 by 3 geometry. Nine parameters are needed for the material identification of
the subcells.
Table 9: Nine-Cell RUC geometry for equivalent constituent volume fractions
System Geometry XH 1
XH 2 XH 3
XL 1
XL 2 XL 3
1
1.
0.06
0.6066758
1.
0.06
0.6066758
0.076369
1.
0.076369
0.70486
1.
0.70486
0.173
1.
0.173
0.53186
1.
0.53186
0.184028
0.184028
0.515228
0.515228
94
Xl
Xh
(1)
Jute Fiber
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
Glass Mat
Figure 36: Nine-Cell RUC geometry for System 1
95
Table 10: Nine-Cell RUC in-plane cell properties for equivalent volume fractions
System Geometry E11 MPa (ksi) E22 MPa (ksi) 12
G12 MPa (ksi)
1
19950 (2893)
5390 (782)
0.27 2390 (347)
19950 (2893)
5100 (739)
0.25 2520 (366)
19930 (2890)
5150 (747)
0.25 2300 (333)
19940 (2892)
4960 (720)
0.25 2540 (369)
96
3.5
3
3 2
2.5
2
0
1
3
3
6 3
4 5
0 10
2
1
3
1
3
3
0.6
0.8
0.4
0.5
0.2
1
2
3
0.5
0
3
1.5
3
3
42
4
3
4
2
5
6
1
12
3
3
140
26 5
1 2
3
3
1.5
2.5
2
2
2
2
3
3
10
10
2
3
1.2
1.4
1.6
0.2
(a) TSA: 1 Hz, 200lb load amplitude
0.4
0.6
0.8
1.2
1.4
1.6
(b) Digital Image Correlation at 400 lb load
Figure 37: Contour map of TSA image scaled to predicted remote axial stress and
contour map of calculated axial stress from an optical strain analysis method of an
open-hole specimen (Specimen 1) of Jute with chopstrand mat surface layer
Figure 37 shows the same open-hole specimen analyzed experimentally using TSA
as shown in Fig. 37.a and with Digital Image Correlation (DIC) as shown in Fig.
37.b. The TSA values are scaled by the remote axial stress and the DIC values are
scaled by the axial modulus to simplify correlation with the proposed computational
models.
97
Figure 37 shows the same open-hole specimen analyzed experimentally using TSA
as shown in Fig. 37.a. The TSA values are scaled by the remote axial stress to simplify
correlation with the proposed computational models.
10
2
2
2
2
0.4
0.6
0.8
2
2
2
0.2
0.5
2
2
2
2
2
2
1
1
2
2
5
410 21
1 0
0
3
2
2
3
3
2
2
140
6
1.5
2
1
2
60 3 152
2
4
2
2
2
2
2
2
2
3
2
2.5
2
3
1.2
1.4
1.6
(a) TSA: 1 Hz, 200lb load amplitude
Figure 38: Contour map of TSA image scaled to predicted remote axial stress and
contour map of calculated axial stress from an Digital Image Correlation method of
an open-hole specimen (Specimen 2) of Jute with chopstrand mat surface layer
98
Figure 39 shows the two macromechanical nonlinear orthotropic computational

models used in this study.
ADT
APT
Figure 39: Contour map of predicted remote axial stress of open-hole specimen Jute
with chopstrand mat surface layer showing ADT and APT Models
Figure 40 shows a linear orthotropic response compared to the Nine-Cell constitutive model and the two macromechanical nonlinear orthotropic computational models
used in this study.
99
Linear
Nine-Cell
Figure 40: Contour map of predicted remote axial stress of open-hole specimen Jute
with chop strand mat surface layer showing Linear and Nine-Cell Models
100
In addition to the qualitative analysis shown for the multi-axial state of stress
modeled in Figures 39 and 40, a line interrogation is shown in Figure 41 to show the
prediction capabilities of the Nine-Cell model.
12
TSA specimen 1
TSA specimen 2
DIC
9 Cell
Linear
10
= 3/8 in
Centered
8
10 in
1.725 in
0
1
0.8
0.6
0.4
0.2
0.2
0.4
0.6
0.8
Figure 41: Axial Stress line interrogation on open-hole specimen for 400lb loading on
Jute with chop strand mat material system showing the prediction capabilities of the
Nine-Cell Model and Linear orthotropic models
101
The prediction capabilities are shown quantitatively for the Nine-Cell model in
Figure 42 using a horizontal line interrogation of the axial stress across the open hole
for a 400 lb loading.
12
TSA specimen 1
TSA specimen 2
DIC
ADT
APT
Linear
10
= 3/8 in
Centered
10 in
6
1.725 in
0
1
0.8
0.6
0.4
0.2
0.2
0.4
0.6
0.8
Figure 42: Axial Stress line interrogation on open-hole specimen for 400lb loading on
Jute with chop strand mat material system showing the prediction capabilities of the
ADT, APT and Linear orthotropic models
102
In addition to the macro and micromechanical models the Extended finite element
fracture modeling framework (XFEM) is applied here to model the pregressive damage
in the biocomposite coupons. The XFEM is calibrated using experimentally derived
fracture parameters. The energy release rate was calculated from the translaminar
fracture toughness tests, and damage initiation principal stresses were taken from
the ultimate uni-axial strength values. Linear orthotropic material properties were
assigned to the elements. Eccentrically Notched Tension specimens manufactured
from Jute/ Glass Chopstrand Mat panels were tested, and the results are shown in
Figure 43.
0.2 W
diameter
0.015 W
max
A
0.2 W
an
1.5 W
1.5 W
B; as
required
4.0 W min
Figure 43: Translaminar Fracture Toughness Test Specimen Schematic

Ultimate properties of the Jute/glass chopstrand mat system are shown in Table
6. To calculate the translaminar fracture toughness ASTM E1922 gives the following
equation: K =
P
1/2 [1.4
BW 1/2
+ ][3.97 10.88 + 26.252 38.93 + 30.154
9.275 ]/[1 ]3/2. Where K = applied stress intensity factor, MPa m1/2, P= applied
load, MN, = a/W (dimensionless), an = notch length, m, B = specimen thickness,
m, and W= specimen width.
The opening mode energy release rate, GI, for a crack under plane stress conditions
in an orthotropic material is related to the mode-I SIF through relations given by Suo
et al. [121]:
103
Test 1
Test 2
Test 3
Test 4
0.5
0.4
W
Force (kips)
0.3
A
0.2 W
diameter
1.5 W
0.015 W
max
A
4.0 W
min
0.2
an
1.5 W
0.1
0.2 W
A
B; as
required
0.05
0.1
0.15
0.2
Crack Mouth Opening Displacement (in)
0.25
0.3
Figure 44: Results of the Translaminar Fracture Toughness Test on Jute/Glass Chopstrand using the Eccentrically Loaded Single Edge Notched Tension Specimen
104
G1 =
1+
s11s22
2
1/2
1/4 KI2
(170)
where sij are the compliances and the two parameters, q and ,are measures of
the in-plane material orthotropy
E2
2s12 + s66
(E1E2 )1/2
s11
=
, =
=
(12 21)1/2
s22
E1
2 s11s22
2G12
(171)
ABAQUS 6.9 allows users to model propagating cracks without prescribing the
crack path with the XFEM framework using a version of the superimposed element
formulation originally introduced by Hansbo et al. [64]. The method was then incorporated into a traditional FEM framework by Song et al. [117]. At this point
the commerically available XFEM framework is limited to linear continuum elements
(CPE4, CPS4, C3D4, C3D8) with or without reduced integration. The framework
does not allow for crack branching, interacting cracks, multiple cracks within an element, or cracks turning more than 90 degrees within an element.
Figure 45 shows the remote axial stress vs. strain at location 1 (Centered horizontally and 1.63 inches above the open hole). The APT Model shows the greatest
nonlinearity, but all of the models are close for the prediction of stress/ strain response at location 1.
105
18
16
DIC
ADT
APT
9 Cell
Linear
XFEM
14
Remote Axial Stress (ksi)
12
10
1.63 in
8
10 in
= 3/8 in
Centered
4
1.725 in
4
Strain (in/in)
7
3
x 10
Figure 45: Remote Stress vs. Remote Axial Strain, Image Correlation at Location
1 used for validation of the ADT, APT, Nine-Cell models, Linear, and XFEM orthotropic models
106
Figure 46 shows the remote axial stress vs. strain at location 2 (Centered vertically and at the edge of the open hole). Again the APT Model shows the greatest
nonlinearity, followed by the ADT and Nine-Cell models. The linear model is shown
here as a control for comparative purposes.
18
DIC (L)
DIC (R)
ADT
APT
9 Cell
Linear
XFEM
16
14
*
@
I
@
12
10
= 3/8 in
Centered
10 in
*Damage passed elements covered

by the strain gauge area.
0.005
0.01
0.015
Strain (in/in)
0.02
1.725 in
0.025
0.03
Figure 46: Remote Stress vs. Axial Strain at Open Hole Edge, Image Correlation
at Location 2 used for validation of the ADT, APT, Nine-Cell models, Linear, and
XFEM orthotropic models
107
Figure 47 shows the remote axial stress vs. strain at location 3 (Centered vertically and 0.625 inches from the center of the open hole). The ADT model shows the
greatest nonlinear behavior followed by the APT and Nine-Cell models. The linear
model is shown here as a control for comparative purposes.
18
DIC
ADT
APT
9 Cell
Linear
XFEM
16
14
12
10
= 3/8 in
Centered
8
10 in
0.685 in
4
1.725 in
0.001
0.002
0.003
0.004
0.005
Strain (in/in)
0.006
0.007
0.008
0.009
0.01
Figure 47: Remote Stress vs. Axial Strain, Digital Image Correlation at Location
3 used for validation of the ADT, APT, Nine-Cell models, Linear, and XFEM orthotropic models
108
Figure 48 shows the remote axial stress vs. strain across the open hole (3/4 in
gauge length extensometer). The ADT model shows the greatest nonlinear behavior
followed by the APT and Nine-Cell models. The linear model is shown here as a
control for comparative purposes.
18
DIC
ADT
APT
9 Cell
Linear
XFEM
16
14
12
10
= 3/8 in
Centered
in
10 in
extensometer
4
1.725 in
0.005
0.01
0.015
0.02
Strain (in/in)
0.025
0.03
0.035
Figure 48: Remote Stress vs. Axial Strain, Digital Strain Analysis using 3/4 inch
gauge length extensometer over the open hole used for validation of the ADT, APT,
Nine-Cell models, Linear, and XFEM orthotropic models
109
CHAPTER IV
THERMOELASTIC STRESS ANALYSIS FOR FAILURE

INITIATION AND PROGRESSION IN COMPOSITE LAP
SHEAR JOINTS
4.1
Introduction
An experimental technique using infrared (IR) thermography is presented to characterize damage initiation and progression in fiber reinforced polymeric (FRP) single
lap shear joints. This experimental work formed the basis for a damage initiation
investigation on single lap joints in the Masters Thesis by Shane Johnson [73] along
with addional experimental work with acoustic emission and clip gauges. The acoustic emission and clip gauge results were taken out, and the TSA results are shown
here as a basic explanation of how damage initiates in a static case. This is essential
because the work is expanded on for experimental fatigue damage analysis using similar methodology, and the same damage progression is expected for single lap joints.
In addition to the experimental work computational fracture models for single lap
joints are developed in the next chapter. Two composite lap joint systems are tested.
The first consists of carbon-reinforced woven and the second is made of laminated
orthotropic plates. The resin layer used for the bonded lap-joints is the FM300K
adhesive with an approximate thickness of 10 mils. Thermoelastic Stress Analysis
(IR-TSA) is proposed to detect expressions of damage initiation at the surface or
the edge of the joint areas. Tests are also conducted with side and direct view IR
measurements detecting the crack tip location and progression with increasing applied mean load. Several samples are removed from loading after initial cracking is
detected and underwent micrographic inspection to characterize the nature of their
110
failure initiation. The proposed IR testing schemes indicate a good ability to detect
expressions of damage initiation in the glue lines near the edge area. Moreover, crack
tip and free surface are also detected as loading progressed, thus IR provides a means
of gauging the severity of damage. The demonstrated IR ability to detect initiation
and progression in lap joints is well suited for future long-term fatigue tests and can be
used to compare the mechanical performance of various FRP joints. This manuscript
is divided into three parts. In the first part, the IR-TSA setup is introduced. The
second part gives motivation for IR-TSA and introduces two techniques for capturing
static damage initiation on lap joints. The third part presents the results of the two
IR-TSA methods and the discussion follows.
4.2
Proposed IR-TSA Testing Set-up
The areas of interest for this study and viewing window for IR images are shown in
Figure 50. A schematic showing the expected IR image results from the front view is
shown in Figure 51. The initial break-up in the adhesive fillet is the initial expected
state of damage; failure initiation can be defined as the formation of discontinuities
in the external bondline of the single lap shear joint. Photomicroscopy will focus on
the areas of discontinuities as detected from infrared thermography to confirm the
IR-TSA damage observation. A DeltaTherm DT1500 thermoelasticity measurement
system [1] was used to acquire the thermal measurements. This system has an infrared
array detector synchronized with the applied cyclical loading in order to measure
the transient thermoelastic effect and filter out the IR emmissions not associated
with the material strain energy releases. The infrared detector acts as a transducer,
which converts the incident radiant energy into electrical signals. A lock-in analyzer
(a signal-processing unit) extracts the thermoelastic information from the detectors
output signal by using the reference signal from the loading device. The TSA-IR
system uses the reference signal to reject any non-stress related thermal emissions.
111
The Delta-Therm has a thermal resolution of approximately 1 mK for image exposure

times of one minute or less. Figure 49 shows the testing setup. The IR camera
captures images at rates of more than 400 frames per second. The applied load
signal is used to integrate synchronized TSA images that correspond to peak values
of loading. The integration of the captured images is a temporal smoothing process
performed over a specified period. In this study, a period of 1-2 min was used.
The cyclic load was applied using an MTS 810servo-hydraulic test system with a 22.2
kN(50 kip) capacity. The accuracy of the recorded strains is within 50 microstrostrains
and the load is within 0.22 kN (50lb).
Figure 49: Infrared camera and the IR-TSA experimental setup
112
Window for
IR-TSA
Lap Joint
+
(a) Side View of Specimen
(b) IR Window
Figure 50: Areas of interest and IR Window
113
(a) Undamaged
(b) Damaged
Figure 51: Schematic for Expected IR Front View of Lap Joint
4.3
Proposed IR Testing Approach
Infrared thermography was applied on composite lap-joints to provide a means of

measuring temperature changes on the surface due to applied cyclic loading. Two
joint types were examined. These include long and short bond geometries made of
carbon-reinforced woven and quasi-isotropic laminated parts. The overall goal was
to characterize damage evolution with an emphasis on detecting failure initiation.
Progression of damage is linked to spatial temperature changes in composite single lap
shear joints which prompts the use of an IR-TSA technique. The IR-TSA technique
shows promise as a refinement over other testing techniques used in detecting failure
initiation and progression in composite single lap shear joints. The motivation of
focusing the IR testing for initiation on the bond line can be explained through a
FE stress analysis of a typical lap joint. This geometry was modeled in the Abaqus
general purpose commercial FE code [2] with 2620 CPE4 plane strain elements. The
ends of the specimen are fixed as noticed by the lack of rotation in Figure 53. The
properties used in the model are listed in Table 11.
114
Table 11: Unidirectional material properties of IM7/epoxy composite laminate

E11 GP a(ksi) E22 GP a(ksi) G12 GP a(ksi) 12
P ly T hickness mm(in)
142.6 (20,700) 8.39 (1220)
3.9 (566)
0.34 .16 ( 0.0063)
1 Inch
8 Plies
2 Inches
6 Inches
Figure 52: Single Lap Joint specimen geometry made from IM7 Quasi-isotropic laminates
115
Figure 53: Deformed single lap joint
The FE analysis of a typical IM7/epoxy quasi-isotropic single lap joint (Figure

67) is used to show that the majority of the shear stress in a single lap shear joint is
confined to a relatively small region near the external bond line (Figure 54) while the
interior bond area is relatively unloaded. There is a stress intensity at the bond edge,
and high peel stresses are the main cause of failure initiation at this location in single
lap joints. A snapshot of the FE results in the form of peel stresses in this loaded
joint as shown in Figure 55 gives insight into the extent of the localized singularity
and the bond edge.
150
100
Peel Stress (MPa)
Shear Stress (MPa)
50
50
100
150
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
Normalized Bond Length

Figure 54: Peel and shear stresses in a typical IM7/Epoxy quasi-isotropic single lap
shear joint
Two IR-Thermography testing methods, termed Method-A and Method-B were
applied in the proposed experiment to detect damage initiation of the lap shear joints.
In Method-A, the load is applied monotonically and paused at select levels for short
durations. During this time, a small cyclic incremental load is applied and coupled
116
Figure 55: Localized stress intensity at bond edge, FE results showing peel stresses
at bond edge on a deformed IM7/Epoxy quasi-isotropic single lap joint
with the IR-TSA measurements (Figure 56). In Method-B, once the applied load has
reached a desired level the loading is reversed to a small mean load for a short duration, in which a cyclic load is applied to be coupled with the IR-TSA measurement.
The process continued on by increasing load magnitudes followed by applied cyclic
loading can be repeated until ultimate failure is reached. Method-A was primarily
used for damage progression while Method-B was used to detect initiation. It should
be mentioned that the later method was time consuming and can be hard to automate
due to the need to process each result to determine that initiation has occurred. The
advantage of Method-B lies in its ability to detect damage at an unloaded state, and
this method prevents additional damage to the joint due to the added cyclic loading.
Failure initiation was defined in this experiment as a stress drop at points on the
external edge of the bond.
117
Force
Force
Time
Time
(a) Method A
(b) Method B
Figure 56: Two proposed IR-TSA Test Methods. Method-A: a static load step is
applied, followed by a cyclic loading about this loading level. Method-B: a static load
is applied and unloaded followed by cyclic loading at a pre-specified load and cyclic
load amplitude. The next applied static load level is increased.
118
4.4
Results
Representative results are presented to examine the proposed testing methods. Initiation in both Method-A and Method-B is defined as a distinct difference and stress
drop in the stress concentration at the bondline. A series of specimens was tested
using Method A. Figure 57 shows the measured TSA response to test Method-A with
a woven long-bond lap-shear joint. It is interesting to note that that the continuous
stress concentration in the external bondline present at a mean load of 200 lb and
how the IR line is diminished with increased loading up to 4,400 lb. Discontinuities
in this continuous line of stress concentration are present in the woven long-bond
specimen as early as 1750 lb. The same damage patterns are seen in lap-joints made
from uni-tape long bond specimen as shown in Figure 58 also tested with Method-A;
however, the crack front is more easily detected by monitoring the side of the uni-tape
long bond specimen.
Figure 57: Front and side view of the crack front formation in long-bond single-lap
shear joint made with woven plates and tested with Method A, ultimate failure occurs
immediately following the 4,450 lb load level. Signature of crack tip is visible from
the side view at the 4,400 lb load level.
119
Figure 58: Front and side view of the crack front formation at the bond edge in
long-bond single-lap shear joint made from uni-tape plates tested with Method A,
ultimate failure occurs immediately following the 3,300 lb load level. Signature of
crack tip is visible from the side view at the 3,200 lb load level.
120
The following series of specimens include short-bond lap joints made of woven and
uni-tape plates shown in Figures 59 and 60 were tested using Method-B. The external
bond in both cases has initial discontinuities and nonuniformity. This is detected from
the early IR-TSA measurments unlike the long-bond case. The continuous stress
concentration seen in Figure 57 above is not always the initial case due to slight
differences in the manufacturing at the bondline. An attempt was made in each
specimen to reach a condition of a discontinuous bondline in order to determine the
early stages of crack formation.
Figure 59: Early stages of crack formation and damage initiation in short-bond singlelap shear joint made from woven plates tested with procedure B preserved after a
maximum load level of 1700lb. Noticeable changes in the bond line IR measurement
from (c) to (d).
Figure 60 demonstrates the difference between the response of the woven short
bond specimen and the uni-tape short bond specimen. The uni-tape short bond
specimens emit relatively very little acoustic or thermal responses to loading compared
to the other specimen types. Figure 8 shows how failure initiation was captured
utilizing TSA.
Several specimens were preserved after noticeable bond line damage initiation has
occurred and taken for further investigation using destructive sectioning coupled with
photo-microscopy. The major assumption in the proposed testing technique is once
a noticeable IR damage detection has occurred in the bond line, damage within the
121
Figure 60: Early stages of crack formation and damage initiation in short-bond single
lap-shear joint made with uni-tape plates tested with Method B and preserved after a
maximum load level of 2000lb. Increasing changes in bondline IR emission indicating
damage progression from (b) to (c) and (c) to (d).
Figure 61: Early stages of crack formation and damage initiation in long-bond singlelap shear joint made from woven plates tested with Method B and preserved after
a maximum load level of 3750lb. Increasing changes in bondline IR emission indicating damage progression from (b) to (c) and (c) to (d). Noticeable drop in stress
concentration from (c) to (d).
joint area should be visually evident. Specimens were removed from testing prior to
ultimate failure once thermography, showed that initial failure may have occurred.
This allowed for marking of the specimens at the locations were the suspected failure
initiation occurred. The objective of the failure analysis was to document, if possible,
the nature and extent of this initial damage. Prior to sectioning the specimens for
micro graphic inspection, each specimen was ultrasonically inspected for evidence of
122
Figure 62: Early stages of crack formation and damage initiation in long-bond singlelap shear joint made from uni-tape plates tested with Method B and preserved after a
maximum load level of 1900lb. Increasing changes in bondline IR emission indicating
damage progression from (c) to (d). Noticeable drop in stress concentration from (c)
to (d).
macroscopic flaws (e.g. delaminations greater than 1/4 diameter). As expected no
evidence of this type of damage could be found at the resolution of the C-scan system
used. Each specimen was then sectioned and potted in preparation for micrographic
inspection. Potted specimens were ground down to points near the marked IR-TSA
initiation flaw points using coarse grit paper. At this point specimens were polished
down to the flaw points while being checked frequently for indications of cracking.
In most cases, evidence of micro-cracking was documented at the sites suspected of
failure. Several examples of photos taken at 50X and 200X magnifications are provided. Figure 63 through 64 show several photo-microscope images that correlate well
with the above assumption and provide a substantial confirmation to the proposed
IR damage initiation and progression method.
123
Figure 63: Photograph showing micro-cracking in short bond line fabric specimen #1
(200X) corresponding to IR-TSA measurement in Figure 59.
Figure 64: Photograph showing disbonding at the adhesive interface in short bondline
fabric specimen #4 (200X)
124
Figure 65: Photograph showing microcracking in short bond line uni-tape specimen
#5 (50X) corresponding to IR-TSA in Figure 60.
125
CHAPTER V
A STOCHASTIC FATIGUE DAMAGE MODEL FOR

COMPOSITE SINGLE LAP SHEAR JOINTS BASED ON
MARKOV CHAINS AND THERMOELASTIC STRESS
ANALYSIS
5.1
Introduction
Quantitative damage detection in composite single lap joints during fatigue is difficult
and not practical with many traditional experimental methods. The objective of this
study is to build on the existing experimental methods previously used for investigations on static damage initation by applying stochastic Markov chain methods to
model the fatigue damage evolution in composite single lap-joints subjected to cyclic
mechanical loadings and monitored in-situ by Thermoelastic Stress Analysis. Thermoelastic stress analysis (TSA) discussed more in detail in the introduction has been
used in various fracture and fatigue studies for a quantitative alternative to deformation measurements, and those more relevent to this study are discussed here. Potter et
al. [104, 105] developed techniques to investigate the thermoelstic effect in laminated
composites. Bakis and Reifsneider [13] used laminate analysis and micromechanical
formulations to develop predictions of the thermoelastic signal, and good agreement
was found between their analytical analysis and experiments, with differences due
possibly to material heterogeneity and laminate surface conditions. They also investigated the effect of the cyclic loading frequency on the thermoelastic singal to
find spurious non-adiabatic effects above 30 Hz. Dulieu-Barton et al. [42] developed
a temperature correction methodology for fiber reinforced polymeric materials and
126
metals to account for a mean ambient temperature effect during TSA measurements.
Their methodology is better able to filter out effects outside of the cyclic loading
such as the operator handling the specimen during acquisition or heat flow from the
grips to the specimen during cyclic loading. Cavaliere et al. [33, 34] applied TSA
techniques to measure crack growth in-situ during fatigue damage while capturing
stress fields translated from thermoelastic stress analysis developed to measure stress
intensity factors (SIFs) for notched metal matrix composite specimens. Tomlinson
and Marsavina [127] and Diaz et al. [39, 40] used Muskhelishvilis solution to calculate SIFs for fracture analysis of metallic materials using TSA techniques providing
methods to determine the crack tip location and residual stresses in their study. ElHajjar and Haj-Ali [48, 63] proposed a TSA calibration technique to quantitatively
map the IR signal to the sum of in-plane strains in orthotropic laminate composites, and the experimental results compared well with a finite element analysis of
thick section composites with an transversly isotropic surface layer. Emery et al. [50]
similarly developed calibration techniques relating the in-plane strains and material
properties to predict the thermoelastic signal in orthotropic laminates taking into
account the surface characteristics. Haj-Ali et al. [62] also developed IR-TSA methods with and without transversely isotropic surface coatings to measure of the sum
of normal in-plane strains to show that mixed mode SIFs can be calculated using
their method and the Lekhnitskiis anisotropic elasticity solution for the stress field
ahead of a crack. TSA techniques were applied by Johnson et al. [74] to investigate the fatigue behavior of pultruded materials to obtain an averaged TSA response
related to fatigue damage evolution. In their study the compliance showed lower sensitivity under the fatigue loading than TSA measurements. Johnson [73] also tested
S2-glass/E733FR open-hole specimens with a [45/0/45/90]s lay-up sequence to verify
that TSA could be related to intermittent compliance measurements and developed
127
to capture cumulative fatigue data in-situ. A limited number of specimens was examined; therefore, it could not be used to predict fatigue damage evolution. This Markov
chain method presented here was developed by Wei et al [132] for analyzing cumulative damage effects in notched laminated composites although similar approaches
can be found in the literature [110, 137]. The temperature correction methodology
developed by Dulieu-Barton [42] is recommended but was not taken into account in
this study because the damage severity in single-lap joints with thin adherends can be
determined by TSA by locating the debonding fronts with acceptable accuracy using
the algorithms discussed in this paper. The material system in combination with
the open-hole specimen geometry chosen for the previous study [132] with open-hole
specimens was difficult to analyze due to a host of competing damage mechanisms
creating non-adiabatic effects that had to be filtered out by post processing algorithms before the cumulative damage model could be calibrated. The choice of single
lap joints allows for a better test of the modeling framework by limiting the analysis
to a single damage mechanism, debonding of the joint adherends. This also simplifies
the experimental analysis and increases the accuracy of the calibration by limiting
the role of non-adiabatic effects in the TSA measurements. Of course, it is impossible
to eliminate the influence of irreversible heating arising from frictional effects on the
delaminated part, but efficient algorithms are developed in this study for analyzing
the joints by decoupling the bonded portion from the delaminated portion where the
non-adiabatic effects emanate from through an efficient set of image analysis algorithms. This paper specifically employs TSA techniques and new post processing
image analysis algorithms during constant amplitude fatigue for uni-directional S2glass/E733FR composite lap-joints to define new damage metrics. The TSA signal
from the surface of S2-glass/E733FR single lap joints is captured continuously during
constant amplitude fatigue loadings. Damage metrics are developed for the composite
samples based on developing delamination fronts as evaluated by the thermoelastic
128
stress analysis (TSA) technique. The TSA damage index at the last fatigue cycle
is used to calibrate the Markov chain models (MCMs). Markov chain models are
employed with the TSA damage index to predict fatigue behavior for a maximum
stress corresponding to 65% and 55% of the ultimate stress. Next, new stochastic
S-N curves utilizing the calibrated Markov damage model are also introduced. The
damage predictions of the MCMs are then examined at different fatigue cycles. A
new method is proposed to construct a predictive stochastic S-N curve utilizing the
MCMs for lap joints. The numerical results from the experiments are utilized along
with Markov chain theories to predict the fatigue behavior accounting for variability
in S2-glass/E733FR lap-joints. The proposed TSA with MCMs is shown to be very
effective in predicting the damage evolution and allowed constructing stochastic S-N
curves for composite single lap shear joints.
5.2
Testing Setup
A DeltaTherm DT1500 thermoelasticity measurement system [1] was used to acquire

the thermal measurements. This system has an infrared array detector synchronized
with the applied cyclical loading in order to measure the transient thermoelastic
effect and filter out the IR emmissions not associated with the material strain energy
releases. The infrared detector acts as a transducer, which converts the incident
radiant energy into electrical signals. A lock-in analyzer (a signal-processing unit)
extracts the thermoelastic information from the detectors output signal by using the
reference signal from the loading device. The TSA-IR system uses the reference signal
to reject any non-stress related thermal emissions. Figure 66 shows the testing setup.
The IR camera captures images at rates of more than 400 frames per second. The
applied load signal is used to integrate synchronized TSA images that correspond
to peak values of loading. The integration of the captured images is a temporal
smoothing process performed over a specified period. In this study, a period of 1
129
min was used and images are acquired every 3 minutes. TheDeltaTherm was set to a
consistent electronic iris of 37%. The cyclic load was applied using an MTS 810servohydraulic test system with a 22.2 kN(50 kip) capacity. The accuracy of the recorded
strains is within 50 microstrostrains and the load is within 0.22 kN (50lb). Out-ofplane displacements are noticeable due to rotation during fatigue loading because of
the relatively high loadings, but in-plane displacements in the bonded area are very
small so there was no need for motion compensation.
Figure 66: Fatigue Test Setup for S2-glass/E733FR Single Lap Joints
130
The proposed IR-TSA technique provides a high resolution full-field thermal map
of the surface of a cyclically loaded material developed here to describe fatigue damage evolution in FRP single lap joints. Fatigue damage, such as delamination, matrix
cracking, fiber debonding, and fiber breaking, can cause stiffness degradation and
spatial stress redistribution on the surface and between the layers. Thus, the thermoelastic signal acquired over the bonded area changes during damage events. Thermoelastic stress analysis of undamaged material relies on reversibility of the load to
obtain adiabatic conditions and increase the accuracy by integrating the IR images.
However, long integration periods with the lock-in filter will cause an averaging of
the TSA signal from damaged areas experiencing friction with the TSA signal from
undamaged area experiencing little non-adiabatic effects.. For the case of monitoring fatigue damage using TSA, it is desired to strike a balance between accuracy
achieved through long integration periods and obtaining TSA integrated images associated with changing damage states. Integrated TSA images of the specimens were
integrated for 1 min and obtained every 3 minutes (900 cycles) during the fatigue
loading using an automatic acquisition process in combination with a reference signal to correlate the raw IR detector data to with the cyclically varying load. The
infrared camera setup is shown in Figure 1. The fatigue test is stopped when the lap
joint separates. The full-field thermal data can then be applied to characterize damage evolution. In this study, fatigue experiments are conducted under two different
applied maximum stresses. Four S-2 glass lap-joint laminates are tested under the
fatigue loading of 65% of their static ultimate stress. A second series of tests consists
of five specimens with same lay-up and geometry is tested under fatigue loadings of
55% of the ultimate stress. The MCMs are used to provide the damage evolution
during fatigue for these two sets of experiments. Nine FRP specimens are cut with
a diamond abrasive wet saw from monolithic 12 inch by 16 inch unidirectional S2glass/E733FR thin panels into 1.0 inch by 10 inch coupons (where a 2 inch overlap
131
is at the mid-length). The joined parts were formed by co-curing, and there is no
indication of fiber waviness in the adherends. The panels are manufactured using
an autoclave following the pre-preg manufacturers specifications (curing cycles). The
lay-up sequence of the specimens is unidirectional and the average (5 repeated tests)
measured elastic properties are listed in Table 12.
Table 12: Unidirectional material properties of S2-glass/epoxy composite laminate
E11 GP a(ksi) E22 GP a(ksi) G12 GP a(ksi) 12 P ly P lyT hickness mm(in)
48.4 (7020)
12.5 (1820)
4.51 (655)
0.3 .22 (0.009)
Typical specimen and dimensions are illustrated in Figure 67. The overall testing
program including the number of specimens tested at different stress levels is shown
in Table 13. The average quasi-static ultimate strength of the single lap joints is 274
MPa or 39.79 ksi. The specimens are sprayed with flat black paint and cyclically
loaded in tension-tension load control with a maximum load corresponding to 65%
of their ultimate strength. The applied cyclic maximum stress is 178 MPa (25.8 ksi)
with a stress ratio of 0.1 and a frequency of 5Hz. The second set of five specimens
follows the same test procedure using the 55% ultimate stress.
1 Inch
8 Plies
2 Inches
6 Inches
Figure 67: S2-glass/E733FR Single Lap Joint specimen geometry
132
5.3
Fatigue damage metrics using IR-TSA
One of the major failure modes observed on the surface of the lap-joint is delamination
resulting from the stress intensity at the debonding front. The thermal surface field is
complicated during damage, especially after delamination because additional thermal
excitations are induced by friction and other non-adiabatic effects. To represent the
effects of delamination, different metrics may be applied to the TSA data as shown
in equation 172. The algorithm for analyzing the TSA images was developed in
MATLAB.
[T SAx, T SAy] = gradient(medfilt2(T SA))
(172)
T SAyf ront = (abs(T SAy) > 100)
(173)
[M, T SAcontiguous] = bwboundaries(T SAyf ront)
(174)
T SAstats = regionprops(T SAcontiguous,0 P ixelList0)
(175)
First, the raw data is accumulated in a particular image in the form of in-phase,
out-of-phase and resultant data so that an objective autophase algorithm can be
perfomed to account for variation of phase over the field of view. The in-phase TSA
image is shown in Figure 4. A median filter is then applied to the in-phase data using
a 3 by 3 neighborhood smoothing process to reduce noise while preserving the edges,
and a one-dimensional numerical gradient of the data is then applied in both the x
and y directions with the default spacing between points of one. Spatial gradients,
such as
T SA
y
(TSAy), can be applied to the image to capture the debonding front
(T SAyf ront), and a threshold can be set to filter out areas with low spatial gradients
Table 13: Number of test specimens and applied stresses

Applied stress level Number of specimens (ID) max MP a(ksi)
65%
4 (1 to 4)
178 (25.8)
55%
5 (5 to 9)
150 (21.8)
133
min MP a(ksi)
17.8 (2.58)
15 (2.18)
as shown in Figure 69 where y represents the axial axis of the specimen in the loading
direction. A matrix of binary values is then acquired through a filtering process that
obtains only the high magnitude of values of the one dimentional gradient in the
y direction, TSAy. The gradient threshold of 100 that obviously depends on the
application and material system here was chosen arbitrarily, but this threshold value
met the requirement of a consistent value that yielded a contiguous region for both the
upper and lower debonding fronts regardless of the intensity of the applied loading.
The one dimensional gradient in the y direction, TSAy, is shown in Figure 69. At this
point any algorithm can be used to separate the upper debonding front from the lower
front for the purpose of calculating damage metrics; however, MATLAB simplifies
the process of locating and defining contiguous regions through a combination of the
bwboundaries and regionprops functions. The upper and lower debonding fronts are
then separate data sets stored within the variable T SAstats. Of course, the debonding
fronts approach each other as the cycles increase, and the specimen never reverts to
a previous damage state. Once the debonding fronts are spatially defined, the area
between the fronts can be used as a damage metric as shown in Figure 70. The
undamaged area A(Ni ) is used to define the damage metric (variable), where Ni is
the fatigue cycle. For comparison with more simple algorithms, the shortest distance
between the debonding fronts can also be taken as a metric for damage as shown in
Figure 71.
134
(a.) 300 Cycles
(b.) 12,200 Cycles
(c.) 25,000 Cycles
(d.) 29,600 Cycles
Figure 68: Typical evolution of maximum IR-TSA spatial gradients for increasing
fatigue cycles
(a.) 300 Cycles
(b.) 12,200 Cycles
(c.) 25,000 Cycles
(d.) 29,600 Cycles
Figure 69: Typical evolution of maximum IR-TSA spatial gradients for increasing
fatigue cycles
135
A = 20,978
(a.) 300 Cycles
A = 15,772
(b.) 12,200 Cycles
A = 12,265
(c.) 25,000 Cycles
A = 8,173
(d.) 29,600 Cycles
Figure 70: Area measures to evaluate the typical damage evolution as defined by
differential IR-TSA under increasing fatigue cycles
136
L=1
L=0.667
L=0.476
L=0.169
(a.) 300 Cycles
(b.) 12,200 Cycles
(c.) 25,000 Cycles
(d.) 29,600 Cycles
Figure 71: Normalized effective bond length measures to evaluate the typical damage
evolution as detected by IR-TSA under increasing fatigue cycles
The damage metric is related to the undamaged/healthy region (gray area in Fig.
70). The latter can be utilized to define the TSA area damage index. The shear lap
region is initially undamaged, and the healthy region decreases as the fatigue cycle
increases. When the decreasing healthy area cannot take the applied stresses, the
lap-joint specimen reaches final fracture. Therefore, the TSA area damage index can
be expressed by:
D(area) (Ni ) = 1
A(Ni)
A(N0)
(176)
where Ni is fatigue cycle, and A(Ni ) is the undamaged area at a given fatigue
cycle, Ni .
Similarly, a different damage index utilizing the difference of debonding fronts can
be defined. The top and bottom debonding fronts, shown in Fig. 71, approach each
other from the edges of the lap joint. The shortest distance between the debonding
fronts decreases due to the fatigue process. The damage index for the difference of
debonding fronts can be defined as:
137
D(dif f ) (Ni ) = 1
L(Ni )
L(N0 )
(177)
where L(Ni ) is the shortest distance between debonding fronts at a given fatigue
cycle, Ni . It is interesting to note that the damage indices, D(area) and D(diff), are
not necessarily one when the specimens fracture, especially for specimens subjected
to higher stress levels. This is because the final fracture occurs within a TSA integration interval, and the TSA method can only capture the last image with the small
remaining undamaged area. However, this dramatic fracture is expected within 900
fatigue cycles maximum (integration interval used in this study). The image more
close to the final failure can be achieved by shortening the integration time if needed.
Similar trends are shown in the damage indices of area and length based metrics
and the process is irreversible. Thus, the damage evolution metric of the specimens
can be used with the Markov chain theory, and their fatigue behaviors can then be
stochastically constructed.
5.4
A cumulative damage model using Markov chain theory
This section introduces a stochastic cumulative damage (CD) method to describe

fatigue damage progression. Previously developed IR-TSA test methods [132] by the
authors have been used in this study. The work of Wei [131] and Rowatt [109] provide
a more thorough review of the Markov Chain Method for cumulative fatigue. The
Markov chain consists of a probability transition matrix (PTM) and distribution. Let
0 be a row vector representing a damage distribution at different damage states for
the Markov chain. The initial damage distribution,{1, 0, ..., 0}, can be expressed as
where all samples are at their undamaged state. The distribution at step n can be
expressed as:
This section introduces a stochastic cumulative damage (CD) method to describe
fatigue damage progression. Previously developed TSA test methods [132] by the
138
authors have been used in this study. The Markov chain consists of a probability
transition matrix (PTM) and distribution. Let be a row vector representing an assumed initial distribution for the Markov chain. The distribution at step n can be
expressed as:
n = 0 P n
(178)
where the P is the PTM. For a CD model, the PTM with b+1 damage states can
be express as:
P =
p0 q0
p1
q1
..
.
0
..
0
.
pb1 qb1
0
(179)
where the ps and qs are conditional probabilities that determine if the current
damage state remains or proceeds to the next state, respectively. If the PTM is
constant through the entire damage process, then the process is termed stationary.
The probability for the first damage state to proceed at the y-th trial is given by:
f(y) = py1 q
(180)
which is a geometric distribution. Therefore, the p and q variables can be obtained

from:
r=
V AR
p
=
;p+ q = 1
EX
q
139
(181)
where EX and VAR are expected value and variance of the geometric distribution,
respectively.
Up to this point, the Markov theory for stochastic processes has been introduced,
and therefore, the fatigue damage evolution through cycles can be constructed once
the required parameters in the Markov chain are calibrated form experiments. The
variance and mean value of the fatigue cycles for a specific damage state or index (e.g.,
failure when the damage index equals one) can be defined. Then, the Markov chain
model shown in Eq. (178) can be constructed and used for describing the fatigue
damage evolution.
5.5
TSA results and Stochastic S-N curves
Figures 72 and 73 show the TSA undamaged area and debonding difference damage
index for the specimens subjected to 65% and 55% stress levels, respectively. Similar
trends are shown in the damage indices of area and length based metrics. Figure
72 shows how the TSA Damage Indices progress towards failure, and the data is
bounded by the Markov Chain Model as described in the previous section. The data
for specimen 4 in the 65% stress level was taken out of Figure 72 because failure
occurs in over 2x104 cycles, and it is more critical to predict a lower bound on the
data.
Table 14 shows the Markov Chain parameters required to calibrate the models in
Figure 72 and 73.
Table 14: Number of test specimens and applied
Applied stress level #Damage States P
55%
6
0.0001818
65%
6
0.0006246
stresses
Q
0.9998182
0.9993754
One of the advantages of the proposed TSA method is that the fatigue process can
be monitored continuously without any interruption due to taking the measurements.
Therefore, it is possible to show through an S-N curve progression how the average
140
0.9
0.8
Lap Joint
TSA Damage Indices
0.7
0.6
0.5
0.4
0.3
sp2 TSA area
sp3 TSA area
sp4 TSA area
sp2 debonding diff.
sp3 debonding diff.
sp4 debonding diff.
Markov Min Bound
Markov Max Bound
0.2
0.1
Stress level = 0.65

R=0.1; Frequency: 5Hz
0
0.2
0.4
0.6
0.8
1.2
1.4
Cycles to failure at the 65% Stress Level
1.6
1.8
2
x 10
Figure 72: Fatigue damage indices of shear lap joints subjected to stress level of 0.65
141
0.9
Lap Joint
0.8
TSA Damage Indices
0.7
0.6
+
0.5
0.4
0.3
0.2
0.1
Stress level = 0.55

R=0.1; Frequency: 5Hz
0
Cycles to failure at the 55% Stress Level
sp5 TSA area

sp6 TSA area
sp7 TSA area
sp8 TSA area
sp9 TSA area
sp5 debonding diff.
sp6 debonding diff.
sp7 debonding diff.
sp8 debonding diff.
sp9 debonding diff.
Markov Min Bound
Markov Max Bound
5
6
x 10
Figure 73: Fatigue damage indices of shear lap joints subjected to stress level of 0.55
142
specimen progresses through fatigue cycles from one damage state to the next at
various stress levels. The S-N curves are constructed based on the two different
tested stress levels. The Minimum Markov Chain Bounds in Figure 11 indicate at
the stress levels tested where 20% delamination, 40% delamination, and failure as
predicted with the model. Also shown in the figure is the average experimental
data for the area TSA data to show the difference between the prediction and the
average measured case. They are fitted to go through the mean cycles to failure of
the specimens subjected to two stress levels. The average shear stress is simply the
ratio of remote force and the area of over lap region. Three different damage indices,
0.4, 0.2, and 0.1, are also selected to generate the progression of S-N curves. The
S-N curves with larger damage indices are expected to have larger fatigue cycles. The
fatigue cycles for the 55% stress level show relatively less shifting in logarithmic scale.
Therefore, the S-N curves exhibit the progression in slope as well.
143
0.8
Min Markov Bound at 20%
Min Markov Bound at 40%
Min Markov Bound at Failure
Mean Exp. at 20%
Mean Exp. at 40%
Mean Exp. at Failure
0.75
0.7
Lap Joint
Normalized Average Shear Stress (ksi)
0.65
0.6
+
0.55
0.5
2
10
10
10
10
Cycles
Figure 74: S-N curve progression for damage index of TSA area
Stochastic S-N curves of the limited number of tested specimens can be constructed utilizing Markov chain theory. Figure 76 illustrates the iterative algorithm
sweeping the entire fatigue cycles to obtain the stochastic S-N curves. We first calibrate the number of damage states, b+1 used in Eq. 179, of the specimens subjected
to the same applied stress level. The mean and variance of the cycles to failure are
applied in Eq. 181, and then the number of damage states in the PTM can be determined by the convergence of the mean cycles to failure from Markov chain prediction
and from tested data. This calibrated number of damage states is used with a mean
cycles to failure function in the MCM. The mean cycles to failure function is a logarithmic function which goes through the mean cycles to failure of the two sets of
specimens of two stress levels. A trial variance, V ar(tr), is used in the calibrated MCM
M CM (AV G)
to obtain the output mean cycles to failure, Nf
144
. If the output mean cycles
0.8
Min Markov Bound at Failure
Max Markov Bound at Failure
Exp. at Failure
Exp. at Failure
0.75
0.7
Lap Joint
Normalized Average Shear Stress (ksi)
0.65
0.6
+
0.55
0.5
3
10
10
Cycles to faliure
10
Figure 75: Constructed stochastic SN data based on two limited data sets and previously calibrated MCM
145
M CM (AV G)
to failure, Nf
does not converges to the input mean cycles to failure, Nf ,
of the current fatigue cycle. The algorithm iterates by increasing the previous trial
variance, V ar(tr), by a small increment, V AR, and then repeat the iteration until
M CM (AV G)
the mean cycles to failure, Nf
converges. Once the iteration has converged,
the trial variance, Var(tr), is calculated based on the previously converged variance.
The procedure is executed through the entire fatigue cycles in order to obtain the
variance or the distribution of the cycles to failure at a different stress level with no
tested data. Therefore, the complete stochastic S-N curve of the tested material is
obtained. The proposed stochastic algorithm for the construction of S-N curves is discussed in Steps 1-3 in Figure 76, and a schematic representation is shown in Figure 77.
146
Step 1. Initialize variables

(0)
INITIALIZE Nf , Nf , V AR(0,0), V AR, Xduty cycle, j = 0,

(f inal)
imax = Nf
/Nf
Step 2. Find convergent MCM variance for each i-step

LOOP i = 1, 2, 3, ..., imax
(i)
(i1)
Nf = Nf
+ Nf
V AR(i,j) = V AR(i1,j)
Step 3. j-iterations for variance within current i-step

j =j+1
V AR(i,j) = V AR(i,j1) + V AR
CALCULAT E p, q USING Eq. 1
INITIALIZE X
(i)
mf = Integer(Nf /X)
LOOP m = 1, 2, 3, ..., mf
X = X + X
(m)
X = 0 P X
M CM (AV G)
CALCULATE Nf
M CM (AV G)
IF |Nf
M CM (M AX)
, Nf
M CM (M IN )
, Nf
(i)
Nf | > T OL,THEN GOTO Step 3
ELSE GOTO Step 2
Figure 76: Proposed algorithm for generating stochastic S-N curves using previously
calibrated MCMs.
147
START
Cycles to
failure from
stiffness/TSA
measurements
Incremental converge
Input mean, Nf, and
trial variance, Var(tr),
into calibrated MCM
(tr)
(tr)
Var =Var +VAR
No
Output a new mean,
NfMCM(AVG) ,
from calibrated MCM
Input mean & variance

into MCM
If |Nf - NfMCM(AVG)|<TOL
Yes
Yes
Nf = Nf+Nf
If N f < Nf
Calibrate
number of
damage state
max
No
END
Figure 77: Schematic representation of the construction of stochastic SN curves.
148
5.6
TSA results and Stochastic S-N curves
This section presents preliminary results on different joint types. This experimental
analysis provides qualitative behavior on the mode of damage progression and limited
quantitative data on delamination and fracture. Further investigation is beyond the
scope of this study. Figures 78, 79, 80, show TSA images of quai-isotropic lay-ups for
an IM7 laminate, single-lap joint, and single nested overlap joint. Figures 8182, 83,
84 show TSA images of quai-isotropic lay-ups for S2-Glass/Epoxy laminate, single
lap joint, and single nested overlap joint during fatigue testing respectively.
Figure 78: Thermoelastic Stress Analysis for Fatigue of an IM7 Laminate
Figure 79: Thermoelastic Stress Analysis for Fatigue of an IM7 Single Lap Joint
149
Figure 80: Thermoelastic Stress Analysis for Fatigue of an IM7 single nested overlap
joint
Figure 81: Thermoelastic Stress Analysis for Fatigue of a Unidirectional S2Glass/Epoxy Laminate
Figure 82: Thermoelastic Stress Analysis for Fatigue of a Quasi-Isotropic S2Glass/Epoxy Laminate
150
Figure 83: Thermoelastic Stress Analysis for Fatigue of a Quasi-Isotropic S2Glass/Epoxy Single Nested Overlap Joint
Figure 84: Thermoelastic Stress Analysis for Fatigue of a Quasi-Isotropic S2Glass/Epoxy Single Nested Overlap Joint
151
CHAPTER VI
CONCLUSION
6.1
Quantitative Thermoelastic Stress Analysis
A experimental thermoelastic stress/strain analysis (TSA) technique was used to investigate multi-axial behavior variety of FRP composites including Carbon/epoxy,
S2-glass/epoxy, and thick-section E-glass/polyester laminates. The new method relates the Thermoelastic signal to the sum of the normal surface strains in FRP composites, and it provides a full-field measurement of the strain invariant that can be
used to verify computational models. If an additional isotropic surface coat is not
applied to the surface of FRP laminates, two calibration constants may be needed
to describe the system under a multi-axial state of stress. Further testing is needed
to provide insight into the effects of the thickness of the isotropic coating and alternative surface coatings in order to guarantee the required adiabatic conditions. In
addition, experimental methods using two calibration constants to eliminate the need
for coating can be the subject of further research.
6.2
Analytical and Experimental Analyses for Jute Hybrid

Bio-Composites
The linear and nonlinear mechanical behavior of Jute hybrid bio-composites was
investigated, and new nonlinear micromechanical and macromechanical modeling approaches are developed for the analysis of the materials. Two of these nonlinear
anisotropic homogeneous models assume that the material is under a state of plane
stress. The third is a nonlinear 3D micromechanical model using a cohesive micromechanical model, a special case (Nine-Cell) of the more general High Fidelity
Generalized Method of Cells (HFGMC) proposed by Aboudi [101] and modified by
152
Haj-Ali and Aboudi [58]. The three models proposed in the analytical study were
able to capture the nonlinearity inherent in the material. The models from are calibrated from uniaxial experiments and validated under a multi-axial state of stress
with Thermoelastic Stress Analysis (TSA) and Digitial Image Correlation (DIC) using full-field stress-strain analysis of plates with a hole. The Nine-Cell model required
linear orthotropic values taken from the literature in the axial and transverse directions to describe the jute yarn in addition to experimental stress vs. strain data in
shear to calibrate the initial shear modulus in all of the consituents and nonlinear parameters for the matrix. This calibration allowed the Nine-Cell model to capture the
nonlinearity in all directions. The Nine-Cell model was validated under uni-axial and
multi-axial states of stress via full-field stress/strain analysis via digital image correlation (DIC) and Thermoelastic Stress Analysis (TSA) for the hybrid bio-composite
system. The Nine-Cell model can be implemented for different fiber/matrix constituents with confidence of predictability with limited set of experimental tests. The
ADT and APT models were able to capture more nonlinearity than the Nine-Cell
model, but the macromechanical models required uni-axial calibrations in the axial,
transverse, and shear modes to capture the nonlinear parameters. While these hybrid composites have relatively low modulus, their material consistency lends them to
be used in many structural applications. Hybrid bio-composites are environmentally
friendly and can provide a sustainable alternative to existing engineering materials in
several applications.
6.3
Damage Initiation of FRP Lap Joints
Critical Stress and VCCT fracture criterion were developed with finite element models
to predict damage initiation in graphite epoxy single lap joints. Two different quasiisotropic lay-ups are compared for the purpose of investigating damage initiation in
the lap joints. The models predict different responses for different lay-ups. The effect
153
of suppressing free-edge delamination on joint strength through re-ordering of the

stacking sequence for eight-plied symmetric quasi-isotropic laminate was investigated
under both monotonic and cyclic loading. A 15% improvement in UTS was observed
for the free-edge delamination suppressed [45/90/ 45/0]s laminates as compared to
the original [0/ 45/90]s laminates. Experiments show that a lay-up with a higher
static strength does not also correspond to a higher fatigue limit. The models show
a greater correlation to the fatigue limit because damage initiation is predicted at a
higher stress level for the [0/ 45/90]s lay-up compared to the [45/90/ 45/0]s .
To further investigate damage initiation in single lap joints, two new experimental
techniques are developed to investigate the static damage initiation using TSA in FRP
lap joints. The experimental results using testing Method-B show that TSA results
and acoustic emission are closely related. An accumulation of acoustic responses is
followed by TSA responses which show changes in the stress concentration at the
bondline. This observation correlates well the finite element results showing that the
majority of the shear stress in single lap shear joints is confined at or near the bondline
while the interior bond of the joint is relatively unloaded. The infrared response at the
bondline measured utilizing TSA in both Methods A and B show that failure initiation
can be defined as the formation of discontinuities in the external bondline of the
single lap joint. The combination of acoustic emission and TSA provides a wealth of
information used for damage detection. Acoustic emission alone provides information
on micro-cracking and the accumulation of those cracks to saturation impending
ultimate failure, but infrared thermography provides a refinement on detecting the
damage severity or a visual inspection of the location of damage. Specimens tested
with Method-B are to be tested destructively focusing on the areas of discontinuities
from infrared thermography. The latter confirm the IR-TSA damage observation.
This study presented an effective new IR-TSA method for detecting damage initiation and progression in single-lap composite joints. The proposed fullfield method
154
was correlated with AE and mechanical load-deformation measurements. Select results were presented for long and short bond area joints as well as fabric versus
quasi-isotropic uni-tape laminates. These results show a promising potential to extend the proposed new testing technique to other joint types and composite structural
components for static and fatigue performance evaluation.
6.4
Stochastic Fatigue of FRP Lap Joints
A new in-situ fatigue degradation monitoring for S2-glass composite laminated lap
joints using TSA techniques is utilized to predict failure. The proposed remote sensing method continuously captures the damage data during the test spatially, and
this is seen as a major improvement on traditional contact methods (mechanical
stiffness/strength measures, radiography, photoelasticity, and acoustic emission for
investigating the fatigue life of FRPs) where the test must be stopped or where the
severity of a damage event cannot be fully captured. The TSA images are post
processed to provide two damage metrics, TSA area and length measures between
debonding fronts within the single lap joint. The TSA area represents the remaining undamaged bonded region providing strength in shear, and the debonding front
difference represents the remaining bonded length in the loading direction. The two
damage metrics show similar trends in the overall fatigue process. In addition, the
continuous monitoring makes the progression of S-N curves possible. Traditionally
S-N curves are defined by final separation. However, the progression of the S-N curves
provides engineers a more convenient approach to define different failure states to predict maintenance scheduling for laminated joints. Where previous fatigue studies on
MCM were used to predict failure at one stress level using traditional contact based
measurements, the proposed MCM requires calibrations from limited number of experiments and can effectively provide the entire stochastic S-N curve for the tested
material systems. The new stochastic S-N curve method using Markov chain theory is
155
validated using two different stress levels for laminated lap joints using non-traditional
test methods. A major goal of this study is to predict the complete stochastic S-N
curve at a given damage state instead of damage evolution under one stress level.
Therefore, the constant conditional probabilities in the PTM are calibrated from the
experimental data directly show sufficient capabilities to construct the entire stochastic S-N curve. Massive fatigue testing efforts can be saved, and satisfactory results
can be obtained by the proposed MCM. Future work may include novel joint types
(such as: the reverse bent, wavy lap, or nested joint types) where fatigue performance
is enhanced but damage progression and resistance becomes more complicated. The
stochastic analysis offered in this study can also be used with failure criteria and
provide reliability for new material/structural systems at various applied load levels.
6.5
Future Work
6.5.1
Fatigue of Lap Joints
Work with Thermoelastic Stress Analysis on joints will continue with an evaluation of
different lay-up sequences and different joint types with new stochastic, computational
and analytical fracture models for delamination in FRP Joints. Limited experimental
analysis of different joint types is shown in Section 5.6. TSA has proven an excellent
tool for in-situ evaluations of fatigue behavior in joints, and differences in behavior of
FRP laminates, single lap joints, single nested overlap joints have been analyzed with
this technique. Damage is accumulated in nested single overlap joints in the adherends
at the joint ends, and failure occurs when the overlap fractures fully followed by
debonding in the joint as seen in the single lap joint.
156
6.5.2
Fatigue of 3D Woven Fiber Reinforced Polymeric Material Systems
Fracture in FRP materials is complicated due to the various competing damage mechanisms, such as matrix cracking, fiber bridging, fiber rupture, fiber pullout, microbuckling, and delamination. More specifically, fracture and damage behavior in FRP
materials is complicated as a result of their inherent heterogeneous and anisotropic
nature which leads to fracture branching (a lack of self similar crack growth) and
crack growth under unstable mixed modes. Fatigue damage further complicates the
design with FRPs for the purpose of service life prediction because of competing
failure mechanisms. A well-known method for characterizing the fatigue behavior in
terms of applied stress-life (S-N) curve is widely adopted for metallic materials as
well as FRP materials. Crack formation and growth is particularly important for
conventional laminates, because cracks can run to and coalesce in the resin rich layers between plies, leading to failure by delamination. However, this failure mode is
not present in 3D woven composites, because unreinforced resin layers do not exist
in them. In addition, there is evidence to suggest that the 3D woven fibers act as
crack stoppers which limit the extent of damage in a cracked composite. Since, 3D
woven composites are insensitive to crack growth could lead to a redefinition of what
really constitutes failure in them. Unfortunately, failure criteria intended to protect
against delamination failure are often applied to 3D woven composites in aerospace
structures in spite of the fact that delamination is not a failure mode for them. In
reality, it is likely that these composites can be operated at higher strains or for
longer design lives without fear of failure, which could lead to weight savings and/or
performance improvements. It has been theorized that it may be more appropriate
to design this type of composite to an S-N curve, similar to metal fatigue, than to a
static strength. This would constitute a radical change in the way composite structures are designed, and it is anticipated that there would be significant resistance to
the approach without substantial experimental verification. It is therefore profitable
157
to develop an innovative computational modeling program using a combination of

probabilistic and deterministic models to account for the average behavior and variability in test results for 3D woven composites. It is necessary to capture spatial
variability in the microstructure with micromechanical models like HFGMC to provide the theoretical motivation necessary on damage mechanisms in 3D composites
(specifically, matrix cracking and fiber bridging) to implement changes in design. It
is also necessary to validate these models with stiffness measurements and innovative
experimental mechanics techniques such as TSA during in-situ fatigue for continuous
full-field expressions of damage in combination with photomicroscopy for enhanced
views of the microstructure for more full verification of the predicted results. The
overall goal of this specific computational and experimental research program expands
the use of composite technology in industry by furthering the use of 3D composites
by theoretically and experimentally verifying the enhanced structural reliability provided by redundant load paths exploited by the material that eliminate delamination
as damage mechanism and provide new failure criteria based on the actual damage
mechanisms exposed by the fatigue of 3D woven composites.
158
REFERENCES
[1] DeltaTherm. Stress Photonics Inc., Madison, WI, 2001.

[2] ABAQUS Theory Manual. Dassault Systmes Simulia Corp., Providence, RI,
USA, 2007.
[3] Aboudi, J. Micromechanical analysis of composites by the method of cells.
Appl. Mech. Rev. 42, 7 (1989), 193221.
[4] Aboudi, J. Mechanics of Composite Materials - A Unified Micromechanical
Approach. Elsevier, Amsterdam, New York, 1991.
[5] Aboudi, J. Micromechanical analysis of fully coupled electro-magneto-thermotlastic multi-phase composites. Smart Mater. Struct. 10 (2001), 867877.
[6] Aboudi, J. The generalized method of cells and high-fidelity generalized
method of cells micromechanical models - a review. Mech. Adv. Materl. Struct
11 (2004), 329366.
[7] Abu-Farsakh, G. New material models for nonlinear stress-strain behaviour
of composite materials. Composites 20 (1989), 349.
[8] Ahmed, K. S., and Vijayarangan, S. Tensile, flexural and interlaminar
shear properties of woven jute and jute-glass fabric reinforced polyester composites. Journal of Materials Processing Technology 207, 1-3 (2008), 330 335.
[9] Akil, H. M., Cheng, L. W., Mohd Ishak, Z. A., Abu Bakar, A., and
Abd Rahman, M. A. Water absorption study on pultruded jute fibre reinforced unsaturated polyester composites. Composites Science and Technology
69, 11-12 (2009), 19421948. doi: DOI: 10.1016/j.compscitech.2009.04.014.
[10] Amijima, S. Adachi, T. Nonlinear stress-strain response of laminated composites. J. Compos. Mater. 13 (1979), 206218.
[11] Arnold, S., Bednarcyk, B., and Aboudi, J. Analysis of internally cooled
structures using a high order theory. Comput. Struct. 82 (2004), 659688.
[12] Avila, A. F., and de O. Bueno, P. An experimental and numerical study
on adhesive joints for composites. Composite structures 64 (2004), 513537.
[13] Bakis, C., and Reifsnider, K. The adiabatic thermoelastic effect in laminated fiber composites. Journal of Composite Materials 25 (1991), 809830.
159
[14] Bakis, C. E., Yih, H. R., Stinchcomb, W. W., and Reifsnider, K. L.

Damage initiation and growth in notched laminates under reversed cyclic loading. Composite Material: Fatigue and Fracture 2 (1989), 6683.
[15] Bank, L. C. Shear properties of pultruded glass frp materials. J. Mater Civil
Eng. 2 (1990), 11822.
[16] Bank, L. C., and Yin, J. Failure of web-angle junction in postbuckled
pultruded I-beams. Journal of Composites for Construction (1999), 17784.
[17] Barone, S., and Patterson, E. Polymer coatings as strain witness in
thermoelasticity. Journal of strain analysis 33, 3 (1998), 223232.
[18] Bednarcyk, B., Aboudi, J., Arnold, S., and Sullivan, R. Analysis of
space shuttle external tank spray-on foam insulation with internal pore pressure.
J. Engrg. Mat. Tech. 130 (2008).
[19] Bednarcyk, B., Arnold, S., Aboudi, J., and Pindera, M.-J. Local
field effects in titanium matrix composites subject to fiber-matrix debonding.
Int. J. Plasticity 20 (2004), 17071737.
[20] Bednarcyk, B., and Yarrington, P. Elasto-plastic analysis of tee joints
using hot-smac. Tech. rep., 2004.
[21] Bisanda, E. T. N. The manufacture of roofing panels from sisal fibre reinforced composites. Journal of Materials Processing Technology 38, 1-2 (1993),
369379. doi: DOI: 10.1016/0924-0136(93)90209-O.
[22] Bisanda, E. T. N., and Ansell, M. P. The effect of silane treatment
on the mechanical and physical properties of sisal-epoxy composites. Composites Science and Technology 41, 2 (1991), 165178. doi: DOI: 10.1016/02663538(91)90026-L.
[23] Bisanda, E. T. N., and Ansell, M. P. J Mater Sci 27 (1992), 16901700.
[24] Bledzki, A., Faruk, O., and Sperber, V. Cars from bio-fibres. Macromolecular Materials and Engineering 291 (2006), 449457.
[25] Bodner, S. Unified Plasticity for Engineering Applications. Kluwer, New
York, 2002.
[26] Bogdanoff, J. L. A new cumulative damage model, part 1. Journal of
Applied Mechanics 45 (1978), 246250.
[27] Bogdanoff, J. L. A new cumulative damage model, part 3. Journal of
Applied Mechanics 45 (1978), 733739.
[28] Bogdanoff, J. L., and Kozin, F. Probabilistic models of cumulative damage. John Wiley & Sons, 1985.
160
[29] Bogdanoff, J. L., and Krieger, W. A new cumulative damage model,

part 2. Journal of Applied Mechanics 45 (1978), 251257.
[30] Borsellino, C., Calabrese, L., Bella, G. D., and Valenza, A. Comparisons of processing and strength properties of two adhesive systems for composite joints. International Journal of Adhesion and Adhesives 27 (2007), 446
457.
[31] Bremond, P., and Potet, P. Lock-in thermography: a tool to analyze and
locate thermomechanical mechanisms in materials and structures. Proceedings
of SPIE, the International Society for Optical Engineering 4360 (2001), 560
566.
[32] Campilho, R. D. S. G., de Moura, M. F. S. F., and Domingues,
J. J. M. S. Modelling single and double-lap repairs on composite materials.
Composites Science and Technology 65, 13 (2005), 1948 1958.
[33] Cavaliere, P., Rossi, G. L., Di Sante, R., and Moretti, M. Thermoelasticity for the evaluation of fatigue behavior of 7005/al2o3/10p metal matrix
composite sheets joined by fsw. International Journal of Fatigue 30, 1 (2008),
198206. doi: DOI: 10.1016/j.ijfatigue.2007.01.021.
[34] Cavaliere, P., Santis, A. D., Panella, F., and Squillace, A. Thermoelasticity and ccd analysis of crack propagation in aa6082 friction stir welded
joints. International Journal of Fatigue 31, 2 (2009), 385392. doi: DOI:
10.1016/j.ijfatigue.2008.07.016.
[35] Chand, N., and Rohatgi, P. Adhesion of previous termsisalnext term fibre
polyester system. Polym. Commun. 27 (1986), 157160.
[36] Cunningham, P., DulieuBarton, J. M., Dutton, A., and Shenoi,
R. Thermoelastic characterisation of damage around a circular hole in a grp
component. Key Engineering Materials 204-205 (2001), 453463.
[37] da Silva, L. F. M., and Adams, R. D. Techniques to reduce the peel
stresses in adhesive joints with composites. International Journal of Adhesion
and Adhesives 27, 3 (2007), 227 235.
[38] Dally, J., and Sanford, R. Strain-gage method for measuring the openingmode stress-intensity factor, ki. Experimental Mechanics 27, 4 (1987), 381388.
[39] Diaz, F. A., Patterson, E. A., Tomlinson, R. A., and Yates, J. R.
Measuring stress intensity factors during fatigue crack growth using thermoelasticity. Fatigue & Fracture of Engineering Materials and Structures 27 (2004),
571583.
[40] Diaz, F. A., Yates, J. R., and Patterson, E. A. Some improvements in
the analysis of fatigue cracks using thermoelasticity. International Journal of
Fatigue 26 (2004), 365376.
161
[41] Dickson, J. N., Hsu, T., and McKinney, J. Development of an understanding of the fatigue phenomena of bonded and bolted joints in advanced
filamentary composite materials. Tech. rep., Lockheed-Georgia Co., 1972.
[42] Dulieu-Barton, J., Emery, T., Quinn, S., and Cunningham, P. A
temperature correction methodology for quantitative thermoelastic stress analysis and damage assessment. Measurement Science and Technology 16 (2006),
16271637.
[43] Dulieu-Smith, J. M., Quinn, S., Shenoi, R. A., Read, P. J. C. L.,
and Moy, S. S. J. Thermoelastic stress analysis of a grp tee joint. Applied
Composite Materials 4, 5 (1997), 283303.
[44] Dunn, S. On the effects of through-thickness thermal conduction on stress
measurement by thermoelastic techniques. Experimental Mechanics 33 (1993),
3236.
[45] Dvorak, G., and Rao, M. Axisymmetric plasticity theory of fibrous composites. Int. J. Eng. Sci. 14 (1976), 361373.
[46] Dweib, M. A., Hu, B., ODonnell, A., Shenton, H. W., and Wool,
R. P. All natural composite sandwich beams for structural applications.
Composite Structures 63, 2 (2004), 147157. doi: DOI: 10.1016/S02638223(03)00143-0.
[47] Edeerozey, A. M. M., Akil, H. M., Azhar, A. B., and Ariffin, M.
I. Z. Chemical modification of kenaf fibers. Materials Letters 61, 10 (2007),
20232025. doi: DOI: 10.1016/j.matlet.2006.08.006.
[48] El-Hajjar, R., and Haj-Ali, R. A quantitative thermoelastic stress analysis
method for pultruded composites. Composites Science and Technology 63, 7
(2003), 967 978.
[49] El-Hajjar, R., and Haj-Ali, R. In-plane shear testing of thick-section
pultruded frp composites using a modified arcan fixture. Composites Part B:
Engineering 35, 5 (2004), 421 428.
[50] Emery, T., Dulieu-Barton, J., Earl, J., and Cunningham, P. A generalised approach to the calibration of orthotropic materials for thermoelastic
stress analysis. Compos Sci Techno 68 (2008), 743752.
[51] Fessel, G., Broughton, J. G., Fellows, N. A., Durodola, J. F., and
Hutchinson, A. R. Evaluation of different lap shear joint geometries for
automotive applications. International Journal of Adhesion and Adhesives 27
(2007), 574583.
[52] Goland, M., and Reissner, E. The stresses in cemented joints. Journal of
Applied Mechanics 11, A17-A27 (1944).
162
[53] Griffin, O., Kamat, M., and Herakovich, C. Threedimensional finite

element analysis of laminated composites. J. Composite Materials 5 (1981),
543560.
[54] Hahn, T. Nonlinear behaviour of laminated composites. J. Compos. Mater 7
(1973), 257271.
[55] Hahn, T., and Tsai, S. Nonlinear elastic behavior of unidirectional composite
laminae. J. Compos. Mater 7 (1973), 102118.
[56] Haj-Ali, R. Cohesive micromechanics: A new approach for progressive damage modeling in laminated composites. International Journal of Damage Mechanics 18, 8 (2008), 691719.
[57] Haj-Ali, R. Nested nonlinear multi-scale frameworks for the analysis of thicksection composite materials and structures, in: Multiscale modeling and simulation of composite materials and structures, 2008.
[58] Haj-Ali, R., and Aboudi, J. Nonlinear micromechanical formulation of the
high fidelity generalized method of cells. International Journal of Solids and
Structures 46, 13 (2009), 2577 2592.
[59] Haj-Ali, R., and Kilic, H. Nonlinear behavior of pultruded frp composites.
Composites Part B: Engineering 33, 3 (2002), 173 191.
[60] Haj-Ali, R., and Kilic, H. Nonlinear constitutive models for pultruded frp
composites. Mechanics of Materials 35, 8 (2003), 791 801.
[61] Haj-Ali, R., Kilic, H., and Muliana, A. Nested nonlinear micromechanical and structural models for the analysis of thick-section composite materials
and structures. Composites Science and Technology 67, 10 (2007), 1993 2004.
[62] Haj-Ali, R., Wei, B.-S., Johnson, S., and El-Hajjar, R. Thermoelastic
and infrared-thermography methods for surface strains in cracked orthotropic
composite materials. Engineering Fracture Mechanics 75, 1 (2008), 58 75.
[63] Haj-Ali, R. M., and El-Hajjar, R. F. A thermoelastic method for strain
measurement in coated orthotropic materials. Experimental Mechanics (2004).
[64] Hansbo, A., and Hansbo, P. A finite element method for the simulation of
strong and weak discontinuities in solid mechanics. computer Method in Applied
Mechanis and Engineering 193 (2004), 35243540.
[65] HART-SMITH, L. Adhesive bonded single-lap joints. Tech. rep., NASA,
1973.
[66] Hart-Smith, L. Analysis and design of advanced composite bonded joints.
Tech. rep., National Aeronautics and Space Administration, 1974.
163
[67] Hashin, Z., Bagchi, D., and Rosen, B. Nonlinear behavior of fiber composite laminates. NASACR-2313 (1974).
[68] Hemelrijck, D. V., Schillemans, L., Roey, F. D., Daerden, I., Boulpaep, F., Wilde, P. D., and Cardon, A. Thermoelastic stress analysis
of fibre reinforced composite systems. 3rd International Conference on Computer Aided Design in Composite Material Technology - CADCOMP 92 (1992),
626633.
[69] Herakovich, C. T., and Mirzadeh, F.
Properties of pultruded
graphite/epoxy. J Reinforced Plastics Compos 10 (1991), 228.
[70] Hong, C., Hwang, I., Kim, N., Park, D., Hwang, B., and Nah. Mechanical properties of silanized jute-polypropylene composites. Journal of Industrial
and Engineering Chemistry 14 (2008), 7176.
[71] Hwang, S. K. C. S. W. Application of a hybrid finite element method to
determine stress intensity factors in unidirectional composites. International
Journal of Fracture 31 (1986), 3751.
[72] Irwin, G. Analysis of stress and strain near the end of a crack traversing a
plate. Journal of Applied Mechanics 24 (1957), 361364.
[73] Johnson, S. Infrafred thermofraphy and thermoelastic stress analysis of composite materials and structural systems. PhD thesis, 2006.
[74] Johnson, S., Kim, H., Wei, B. S., and Haj-Ali, R. Infrared thermography
for damage and fatigue of thick-section frp composites.
[75] Jones, R., Heller, M., Lombardo, D., Dunn, S., Paul, J., and Saunders, D. Thermoelastic assessment of damage growth in composites. Composite Structures 12, 4 (1989), 291314. doi: DOI: 10.1016/0263-8223(89)90077-9.
[76] Jones, R., and Morgan, H. Analysis of nonlinear stress-strain behavior of
fiber reinforced materials. AIAA 15 (1977), 16691676.
[77] Jones, R., and Nelson Jr., D. A new material model for the nonlinear
biaxial behavior of atj-s graphite. J. Compos. Mater. 9 (1974), 1027.
[78] Ju, S., and Rowlands, R. Thermoelastic determination of ki and kii in
an orthotropic graphite-epoxy composite. Journal of Composite Materials 37
(2003), 20112025.
[79] Kageyama, K., Ueki, K., and Kikuchi, M. Fatigue damage analysis of
notched carbon/epoxy laminates by thermoelastic emission and three dimensional finite element methods. Proc. 7th Conf. Comp. Mat. ICCM 7 (1989).
[80] Khan, A., and Huang, S. Continuum Theory of Plasticity. Wiley, New
York, 1995.
164
[81] Kilic, H., and Haj-Ali, R. Elastic-degrading analysis of pultruded composite structures. Composite Structures 60, 1 (2003), 43 55.
[82] Kilic, H., and Haj-Ali, R. Progressive damage and nonlinear analysis of
pultruded composite structures. Composites Part B: Engineering 34, 3 (2003),
235 250.
[83] Kozin, F., and Bogdanoff, J. L. A critical analysis of some probabilistic
models of fatigue crack growth. Engineering Fracture Mechanics 14 (1980),
5989.
[84] Krishnamurthy, W. S. T. Probabilistic and possibilistic analyses of the
strength of a bonded joint. Tech. rep., American Institute of Aeronautics and
Astronautics, 2001.
[85] Krueger, R. The virtual crack closure technique: history, approach and
applications. Tech. rep., 2002.
[86] Kuppuswamy, T., Nada, A., and Reddy, J. Materially nonlinear analysis
of laminated composite plates. Compos. Struct. 2 (1984), 315328.
[87] Kyriakopoulos, M. K., Wilde, W. P. D., Hemelrijck, D. V., and
Schillemans, L. A finite element approach for thermal-structural response in
fibre reinforced composite systems. 3rd International Conference on Computer
Aided Design in Composite Material Technology - CADCOMP 92 (1992), 599
614.
[88] Lekhnitskii, S. Theory of Elasticity of an Anisotropic Body. Holden-Day,
San Francisco, 1963.
[89] Liu, Y., and Mahadevan, S. Stochastic fatigue damage modeling under
variable amplitude loading. International Journal of Fatigue 29 (2007), 1149
1161.
[90] Loeve, M. Probability theory. Springer, New York, 1977.
[91] Lohr, D. T., Enke, N. F., and Sandor, B. I. Analysis of fatigue damage
evolution by differential infrared thermography. Proceedings of the 1987 SEM
Fall Conference (1987).
[92] Mackenzie, A. K. Effects of surface coatings on infrared measurements of
thermoelastic responses. SPIE 1084 (1989), 5971.
[93] Mackin, T. J., and Roberts, M. C. Evaluation of damage evolution in
ceramic-matrix composites using thermoelastic stress analysis. J. Am. Ceram.
Soc. 83 (2000), 337343.
[94] Mahadevan, Y. L. S. Probabilistic fatigue life prediction of multidirectional
composite laminates. Composite Structures 69 (2005), 1119.
165
[95] Mandell, J., and Samborsky, D. Contractor report sand97-3002. Tech.

rep., 1997.
[96] Mathison, S., Pindera, M., and Herakovich, C. Nonlinear response
of resin matrix laminates using endochronic theory. J. Eng. Mat. Tech. 113
(1991), 449455.
[97] Mathur, V. K. Composite materials from local resources. Construction and
Building Materials 20 (2006), 470477.
[98] Muskhelishvili, N. Some Basic Problems of the Mathematical Theory of
Elasticity. P. Noordhoff, Ltd, Groningen-Holland, 1954.
[99] Nahas, M. Analysis of non-linear stress-strain response of laminated fibrereinforced composites. Fibre Sci. Tech. 20 (1984), 297313.
[100] Pagano, N. J., and . Exact moduli of anisotropic laminates. Mechanics of
composite materials (1974), 2344.
[101] Paley, M., and Aboudi, J. Micromechanical analysis of composites by the
generalized cells model. Mechanics of Materials 14, 2 (1992), 127 139.
[102] Petit, P. H. A simplified method of determining the in-plane shear stressstrain response of unidirectional composites. astm stp 460. philadelphia, pa;
1969. ASTM STP (1969).
[103] Pindera, M., and Herakovich, C. An endochronic model for the response
of unidirectional composites under off-axis tensile load. In IUTAM Symposium
on Mechanics of Composite Materials (New York, 1983), Mechanics of Composite Materials: Recent Advances, Pergamon Press, pp. 367381.
[104] Potter, R. Stress analysis in laminated fibre composites by thermoelastic
emission. In SPIE (1987), vol. 731, pp. 110120.
[105] Potter, R., and Greaves, L. The application of thermoelastic stress analysis technique to fibre composites. SPIE 817 (1988), 134146.
[106] Richard, R., and Abbott, B. Versatile elastic-plastic stress-strain formula.
ASCE J. Eng. Mech. Div., Technical Note 101 (1975), 511515.
[107] Roe, P. J., and Ansell, M. P. J Mater Sci 20 (1985), 40154020.
[108] Roh, Y. S., and Xi, Y. A general formulation for transition probabilistic of
markov model and the application of fracture of composite materials. Probabilistic Engineering Mechanics 15 (2000), 241250.
[109] Rowatt, J. D. A probabilistic model for fatigue damage accumulation in
composite laminates. M.s., 1993.
166
[110] Rowatt, J. D., and Spanos, P. D. Markov chain models for life prediction
of composite laminates. Structural Safety 20 (1998), 117135.
[111] Rybicki, E., and Kanninen, M. F. A finite element calculation of stress
intensity factors by a modified crack closure integral. Engineering Fracture
Mechanics 9 (1977), 931938.
[112] Sandhu, R. Non-linear behaviour of unidirectional and angle-ply laminates.
J. Aircraft 13 (1976), 104111.
[113] Shan, A. N., and Lakkard, S. C. Effect of interface modification on the
mechanical properties of polystyrene-sisal fiber composites. Fibre Sci Technol
15 (1981), 4146.
[114] Shukla, A., Agarwal, B., and Bhushan, H. Determination of stress
intnsity factor in orthotropic composite materials using strain gages. Engineering Fracture Mechanics 32 (1989), 469477.
[115] Sih, G., Paris, P., and Irwin, G. On cracks in rectilinearly anisotropic
bodies. International Journal of Fracture Mechanics 1 (1965), 189203.
[116] Simo, J., and T.J.R., H. Computational Inelasticity. Springer, New York,
1998.
[117] Song, J., Areias, P., and Belytschko, T. A method for dynamic crack
and shear band propagation with phantom nodes. International Journal for
Numerical Methods in Engineering 67 (2006), 868893.
[118] Sridhar, M. K., and Basavarajappa, G. Indian J Text Res 79 (1982),
8792.
[119] Sun, C., and Chen, J. A simple flow rule for characterizing nonlinear behavior of fiber composites. J. Compos. Mater. 23 (1989.), 10091020.
[120] Sun, R. W. X. Bio-Based Polymers and Composites. Elsevier Science and
Technology Books, 2005.
[121] Suo, Z., Bao, G., Fan, B., and Wang, T. C. Orthotropy rescaling and
implications for fracture in composites. International Journal of Solids and
Structures 28, 2 (1991), 235248. doi: DOI: 10.1016/0020-7683(91)90208-W.
[122] Tamuzs, V., Dzelzitis, K., and Reifsnider, K. Fatigue of woven composite laminates in off-axis loading i. the mastercurves. Applied Composite
Materials 11 (2004), 259279.
[123] Tamuzs, V., Dzelzitis, K., and Reifsnider, K. Fatigue of woven composite laminates in off-axis loading ii. prediction of the cycle durability. Applied
Composite Materials 11 (2004), 281293.
167
[124] Tanimoto, T. Correlation of statistical fatigue properties in a wide stress

range for frp laminates with failure mechanisms. International Journal of Fatigue 28 (2006), 14931500.
[125] Thomson, W. On the thermoelastic, thermomagnetic and pyro-electric properties of matter. Phil. Mag. 5 (1878), 427.
[126] Tomblin, J., and Barbero, E. J. Compression strength of pultruded fiber
reinforced composites. vol. 1, p. 747. Proceedings of the 10th conference on
engineering mechanics.
[127] Tomlinson, R. A., and Marsavina, L. Thermoelastic investigations for
fatigue life assessment. Experimental Mechanics 44, 5 (2004), 487494.
[128] Van de Velde, K., and Kiekens, P. Thermoplastic pultrusion of natural
fibre reinforced composites. Composite Structures 54, 2-3, 355360. doi: DOI:
10.1016/S0263-8223(01)00110-6.
[129] Varma, I. K., Anantha Krishnan, S. R., and Krishnamoorthy, S.
Composites of glass/modified jute fabric and unsaturated polyester resin. Composites 20, 4 (1989), 383388. doi: DOI: 10.1016/0010-4361(89)90664-2.
[130] Wang, Y., and Zureick, A.-H. Characterization of the longitudinal tensile behavior of pultruded i-shape structural members using coupon specimens.
Composite Structures 29, 4 (1994), 463 472.
[131] Wei, B. S. Thermoelastic Stress Analysis Techniques for Mixed Mode Fracture
and Stochastic Fatigue of Composite Materials. Phd, 2008.
[132] Wei, B.-S., Johnson, S., and Haj-Ali, R. A stochastic fatigue damage
method for composite materials based on markov chains and infrared thermography. International Journal of Fatigue 32, 2 (2009), 350360.
[133] Welch, C., and Zickel, M. Thermal coating characterization using thermoelasticity, vol. 12 of Review of Progress in Quantitative Nondestructive Evaluation. Planum Press, New York, NY, 1993.
[134] Westergaard, H. Bearing pressures and cracks. Journal of Applied Mechanics 6 (1939), 4953.
[135] Wong, A., Jones, R., and Sparrow, J. Thermoelastic constant or thermoelastic parameter? J Phys Chem Solids 48, 8 (1987), 74953.
[136] Wu, W., and Ni, C. Probabilistic models of fatigue crack propagation and
their experimental verification. Probabilistic Engineering Mechanics 19 (2004),
247257.
[137] Xi, Y., and Bazant, Z. P. Random growth of crack with r-curve: Markov
process model. Engineering Fracture Mechanics 57, 6 (1997), 593608.
168
[138] Zadorecki, P., and Flodin, P. Surface modification of cellulose fibers. ii.
the effect of cellulose fiber treatment on the performance of cellulose-polyester
composites. J. Appl. Polym. Sci. (1985), 39713983.
[139] Zeng, Q., and Sun, C. T. Fatigue performance of a bonded wavy compsotie
lap joint. Fatigue and Fracture of Engineering Materials and Structures 27
(2004), 413422.
[140] Zhang, D., Enke, N., and Sandor, B. Thermographic stress analysis of
composite materials. Experimental Mechanics 30, 1 (1990), 6873.
[141] Zhang, D., and Sandor, B. Thermographic analysis of stress concentrations
in a composite. Experimental Mechanics 29, 2 (1989), 1215.
[142] Zhang, D., and Sandor, B. A thermoelasticity theory for damage in
anisotropic materials. Fatigue and Fracture of Engineering Materials and Structures 13, 5 (1990), 497509.
169

Johnson Shane M 201008 PHD

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Johnson Shane M 201008 PHD

Uploaded by

Copyright:

Available Formats

COMPUTATIONAL MODELING, STOCHASTIC AND

EXPERIMENTAL ANALYSIS WITH THERMOELASTIC

Georgia Institute of Technology

COMPUTATIONAL MODELING, STOCHASTIC AND

Professor Abdul-Hamid Zureick

Professor Kenneth Will

Professor Andrew Makeev

Professor Donald White

Date Approved: April 2010

LIST OF SYMBOLS OR ABBREVIATIONS . . . . . . . . . . . . . . . . . .

Fiber Reinforced Polymeric Materials . . . . . . . . . . . . .

1.2 Experimental Studies . . . . . . . . . . . . . . . . . . . . . . . . . .

Quantitative Thermography and Thermoelastic Stress Analysis in Composites . . . . . . . . . . . . . . . . . . . . . . .

TSA for Damage in FRP Composites . . . . . . . . . . . . .

Experimental Studies on Jute Hybrid Bio-Composites . . .

Damage in FRP Lap Joints . . . . . . . . . . . . . . . . . .

1.3 Analytical and Numerical Studies . . . . . . . . . . . . . . . . . . .

Nonlinear Anisotropic Micro-macromechanical Modeling . .

Analytical and Computational Modeling of FRP Lap Joints

Fatigue Damage Models for FRP Composites . . . . . . . .

1.5 Present Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

THERMOELASTIC STRESS ANALYSIS FOR QUANTITATIVE ANALYSIS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

2.3 IR Test Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2.4 Validation of TSA Technique on Prepreg and Pultruded Composites

2.5 Thermomechanical Calibrations for Various FRP Composites . . .

2.6 Validation under a Multi-axial state of stress . . . . . . . . . . . .

MECHANICAL BEHAVIOR OF JUTE HYBRID BIO-COMPOSITES

3.2 Jute Hybrid Bio-composite Systems . . . . . . . . . . . . . . . . . .

3.3 Nonlinear Constitutive Models for Bio-Composites . . . . . . . . .

Anisotropic Deformation Theory (ADT) Based Model . . .

Anisotropic Potential Theory Model (APT) . . . . . . . . .

New Formulation for the Nine-Cell Micromechanical Model

The HFGMC micromechanical modeling framework . . . . .

3.4 Calibration and Verification . . . . . . . . . . . . . . . . . . . . . .

THERMOELASTIC STRESS ANALYSIS FOR FAILURE INITIATION

4.2 Proposed IR-TSA Testing Set-up . . . . . . . . . . . . . . . . . . . 111

5.2 Testing Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129

6.1 Quantitative Thermoelastic Stress Analysis . . . . . . . . . . . . . 152

Fatigue of Lap Joints . . . . . . . . . . . . . . . . . . . . . . 156

Fatigue of 3D Woven Fiber Reinforced Polymeric Material

Effective Elastic Properties . . . . . . . . . . . . . . . . . . . . . . . .

Description of Jute hybrid bio-composite material system . . . . . . .

Nominal V-notch Specimen Dimensions . . . . . . . . . . . . . . . . .

Unidirectional elastic material properties of Jute with chopstrand mat

Calibrated Ramberg-Osgood Relations and Parameters for Jute / Glass

Ultimate material properties of Jute with chopstrand mat . . . . . . .

Fiber linear elasetic properties used in the Nine-Cell RUC . . . . . .

Nine-Cell RUC Isotropic Nonlinear Matrix Material Properties (RambergOsgood Relations) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94

Nine-Cell RUC geometry for equivalent constituent volume fractions .

Nine-Cell RUC in-plane cell properties for equivalent volume fractions

Unidirectional material properties of IM7/epoxy composite laminate . 115

Unidirectional material properties of S2-glass/epoxy composite laminate132

Number of test specimens and applied stresses . . . . . . . . . . . . . 133

Number of test specimens and applied stresses . . . . . . . . . . . . . 140

Schematic cross-sectional view of a layered orthotropic composite with

Schematic of TSA Setup for thermomechanical calibration . . . . . .

Setup for Thermo-mechanical calibration . . . . . . . . . . . . . . . .

Thermo-mechanical calibration for E-glass/epoxy . . . . . . . . . . .

Thermo-mechanical calibration for Uncoated Carbon/Epoxy . . . . .

Schematic showing the lay-up for [05 /90/05 ] . . . . . . . . . . . . . .

Thermo-mechanical calibration for uncoated S2glass/Epoxy for the

Thermo-mechanical calibration for coated S2-glass/epoxy . . . . . . .

Thermo-mechanical calibration for quasi isotropic S2-glass/Epoxy . .